bibtex_url
stringlengths 41
53
| proceedings
stringlengths 38
50
| bibtext
stringlengths 528
3.02k
| abstract
stringlengths 17
2.35k
| authors
sequencelengths 1
44
| title
stringlengths 18
190
| id
stringlengths 7
19
| arxiv_id
stringlengths 0
10
| GitHub
sequencelengths 1
1
| paper_page
stringclasses 528
values | n_linked_authors
int64 -1
15
| upvotes
int64 -1
77
| num_comments
int64 -1
10
| n_authors
int64 -1
52
| Models
sequencelengths 0
100
| Datasets
sequencelengths 0
15
| Spaces
sequencelengths 0
46
| paper_page_exists_pre_conf
int64 0
1
| type
stringclasses 2
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
https://aclanthology.org/2023.findings-emnlp.420.bib | https://aclanthology.org/2023.findings-emnlp.420/ | @inproceedings{litterer-etal-2023-rains,
title = "When it Rains, it Pours: Modeling Media Storms and the News Ecosystem",
author = "Litterer, Benjamin and
Jurgens, David and
Card, Dallas",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.420",
doi = "10.18653/v1/2023.findings-emnlp.420",
pages = "6346--6361",
abstract = "Most events in the world receive at most brief coverage by the news media. Occasionally, however, an event will trigger a media storm, with voluminous and widespread coverage lasting for weeks instead of days. In this work, we develop and apply a pairwise article similarity model, allowing us to identify story clusters in corpora covering local and national online news, and thereby create a comprehensive corpus of media storms over a nearly two year period. Using this corpus, we investigate media storms at a new level of granularity, allowing us to validate claims about storm evolution and topical distribution, and provide empirical support for previously hypothesized patterns of influence of storms on media coverage and intermedia agenda setting.",
}
| Most events in the world receive at most brief coverage by the news media. Occasionally, however, an event will trigger a media storm, with voluminous and widespread coverage lasting for weeks instead of days. In this work, we develop and apply a pairwise article similarity model, allowing us to identify story clusters in corpora covering local and national online news, and thereby create a comprehensive corpus of media storms over a nearly two year period. Using this corpus, we investigate media storms at a new level of granularity, allowing us to validate claims about storm evolution and topical distribution, and provide empirical support for previously hypothesized patterns of influence of storms on media coverage and intermedia agenda setting. | [
"Litterer, Benjamin",
"Jurgens, David",
"Card, Dallas"
] | When it Rains, it Pours: Modeling Media Storms and the News Ecosystem | findings-emnlp.420 | 2312.02118 | [
"https://github.com/blitt2018/mediastorms"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.421.bib | https://aclanthology.org/2023.findings-emnlp.421/ | @inproceedings{li-etal-2023-intra,
title = "Intra-Event and Inter-Event Dependency-Aware Graph Network for Event Argument Extraction",
author = "Li, Hao and
Cao, Yanan and
Ren, Yubing and
Fang, Fang and
Zhang, Lanxue and
Li, Yingjie and
Wang, Shi",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.421",
doi = "10.18653/v1/2023.findings-emnlp.421",
pages = "6362--6372",
abstract = "Event argument extraction is critical to various natural language processing tasks for providing structured information. Existing works usually extract the event arguments one by one, and mostly neglect to build dependency information among event argument roles, especially from the perspective of event structure. Such an approach hinders the model from learning the interactions between different roles. In this paper, we raise our research question: How to adequately model dependencies between different roles for better performance? To this end, we propose an intra-event and inter-event dependency-aware graph network, which uses the event structure as the fundamental unit to construct dependencies between roles. Specifically, we first utilize the dense intra-event graph to construct role dependencies within events, and then construct dependencies between events by retrieving similar events of the current event through the retrieval module. To further optimize dependency information and event representation, we propose a dependency interaction module and two auxiliary tasks to improve the extraction ability of the model in different scenarios. Experimental results on the ACE05, RAMS, and WikiEvents datasets show the great advantages of our proposed approach.",
}
| Event argument extraction is critical to various natural language processing tasks for providing structured information. Existing works usually extract the event arguments one by one, and mostly neglect to build dependency information among event argument roles, especially from the perspective of event structure. Such an approach hinders the model from learning the interactions between different roles. In this paper, we raise our research question: How to adequately model dependencies between different roles for better performance? To this end, we propose an intra-event and inter-event dependency-aware graph network, which uses the event structure as the fundamental unit to construct dependencies between roles. Specifically, we first utilize the dense intra-event graph to construct role dependencies within events, and then construct dependencies between events by retrieving similar events of the current event through the retrieval module. To further optimize dependency information and event representation, we propose a dependency interaction module and two auxiliary tasks to improve the extraction ability of the model in different scenarios. Experimental results on the ACE05, RAMS, and WikiEvents datasets show the great advantages of our proposed approach. | [
"Li, Hao",
"Cao, Yanan",
"Ren, Yubing",
"Fang, Fang",
"Zhang, Lanxue",
"Li, Yingjie",
"Wang, Shi"
] | Intra-Event and Inter-Event Dependency-Aware Graph Network for Event Argument Extraction | findings-emnlp.421 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.422.bib | https://aclanthology.org/2023.findings-emnlp.422/ | @inproceedings{zhang-etal-2023-relevance,
title = "From Relevance to Utility: Evidence Retrieval with Feedback for Fact Verification",
author = "Zhang, Hengran and
Zhang, Ruqing and
Guo, Jiafeng and
de Rijke, Maarten and
Fan, Yixing and
Cheng, Xueqi",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.422",
doi = "10.18653/v1/2023.findings-emnlp.422",
pages = "6373--6384",
abstract = "Retrieval-enhanced methods have become a primary approach in fact verification (FV); it requires reasoning over multiple retrieved pieces of evidence to verify the integrity of a claim. To retrieve evidence, existing work often employs off-the-shelf retrieval models whose design is based on the probability ranking principle. We argue that, rather than relevance, for FV we need to focus on the utility that a claim verifier derives from the retrieved evidence. We introduce the $\textbf{feedback-based evidence retriever} (FER)$ that optimizes the evidence retrieval process by incorporating feedback from the claim verifier. As a feedback signal we use the divergence in utility between how effectively the verifier utilizes the retrieved evidence and the ground-truth evidence to produce the final claim label. Empirical studies demonstrate the superiority of FER over prevailing baselines.",
}
| Retrieval-enhanced methods have become a primary approach in fact verification (FV); it requires reasoning over multiple retrieved pieces of evidence to verify the integrity of a claim. To retrieve evidence, existing work often employs off-the-shelf retrieval models whose design is based on the probability ranking principle. We argue that, rather than relevance, for FV we need to focus on the utility that a claim verifier derives from the retrieved evidence. We introduce the $\textbf{feedback-based evidence retriever} (FER)$ that optimizes the evidence retrieval process by incorporating feedback from the claim verifier. As a feedback signal we use the divergence in utility between how effectively the verifier utilizes the retrieved evidence and the ground-truth evidence to produce the final claim label. Empirical studies demonstrate the superiority of FER over prevailing baselines. | [
"Zhang, Hengran",
"Zhang, Ruqing",
"Guo, Jiafeng",
"de Rijke, Maarten",
"Fan, Yixing",
"Cheng, Xueqi"
] | From Relevance to Utility: Evidence Retrieval with Feedback for Fact Verification | findings-emnlp.422 | 2310.11675 | [
"https://github.com/ict-bigdatalab/fer"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.423.bib | https://aclanthology.org/2023.findings-emnlp.423/ | @inproceedings{lin-etal-2023-train,
title = "How to Train Your Dragon: Diverse Augmentation Towards Generalizable Dense Retrieval",
author = "Lin, Sheng-Chieh and
Asai, Akari and
Li, Minghan and
Oguz, Barlas and
Lin, Jimmy and
Mehdad, Yashar and
Yih, Wen-tau and
Chen, Xilun",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.423",
doi = "10.18653/v1/2023.findings-emnlp.423",
pages = "6385--6400",
abstract = "Various techniques have been developed in recent years to improve dense retrieval (DR), such as unsupervised contrastive learning and pseudo-query generation. Existing DRs, however, often suffer from effectiveness tradeoffs between supervised and zero-shot retrieval, which some argue was due to the limited model capacity. We contradict this hypothesis and show that a generalizable DR can be trained to achieve high accuracy in both supervised and zero-shot retrieval without increasing model size. In particular, we systematically examine the contrastive learning of DRs, under the framework of Data Augmentation (DA). Our study shows that common DA practices such as query augmentation with generative models and pseudo-relevance label creation using a cross-encoder, are often inefficient and sub-optimal. We hence propose a new DA approach with diverse queries and sources of supervision to progressively train a generalizable DR. As a result, DRAGON, our Dense Retriever trained with diverse AuGmentatiON, is the first BERT-base-sized DR to achieve state-of-the-art effectiveness in both supervised and zero-shot evaluations and even competes with models using more complex late interaction.",
}
| Various techniques have been developed in recent years to improve dense retrieval (DR), such as unsupervised contrastive learning and pseudo-query generation. Existing DRs, however, often suffer from effectiveness tradeoffs between supervised and zero-shot retrieval, which some argue was due to the limited model capacity. We contradict this hypothesis and show that a generalizable DR can be trained to achieve high accuracy in both supervised and zero-shot retrieval without increasing model size. In particular, we systematically examine the contrastive learning of DRs, under the framework of Data Augmentation (DA). Our study shows that common DA practices such as query augmentation with generative models and pseudo-relevance label creation using a cross-encoder, are often inefficient and sub-optimal. We hence propose a new DA approach with diverse queries and sources of supervision to progressively train a generalizable DR. As a result, DRAGON, our Dense Retriever trained with diverse AuGmentatiON, is the first BERT-base-sized DR to achieve state-of-the-art effectiveness in both supervised and zero-shot evaluations and even competes with models using more complex late interaction. | [
"Lin, Sheng-Chieh",
"Asai, Akari",
"Li, Minghan",
"Oguz, Barlas",
"Lin, Jimmy",
"Mehdad, Yashar",
"Yih, Wen-tau",
"Chen, Xilun"
] | How to Train Your Dragon: Diverse Augmentation Towards Generalizable Dense Retrieval | findings-emnlp.423 | 2302.07452 | [
"https://github.com/facebookresearch/dpr-scale"
] | https://huggingface.co/papers/2302.07452 | 1 | 1 | 0 | 8 | [
"nvidia/dragon-multiturn-query-encoder",
"facebook/dragon-plus-context-encoder",
"nvidia/dragon-multiturn-context-encoder",
"facebook/dragon-plus-query-encoder",
"facebook/dragon-roberta-query-encoder",
"facebook/dragon-roberta-context-encoder"
] | [
"UKPLab/dapr"
] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.424.bib | https://aclanthology.org/2023.findings-emnlp.424/ | @inproceedings{haraguchi-etal-2023-discovering,
title = "Discovering Highly Influential Shortcut Reasoning: An Automated Template-Free Approach",
author = "Haraguchi, Daichi and
Shirai, Kiyoaki and
Inoue, Naoya and
Kertkeidkachorn, Natthawut",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.424",
doi = "10.18653/v1/2023.findings-emnlp.424",
pages = "6401--6407",
abstract = "Shortcut reasoning is an irrational process of inference, which degrades the robustness of an NLP model. While a number of previous work has tackled the identification of shortcut reasoning, there are still two major limitations: (i) a method for quantifying the severity of the discovered shortcut reasoning is not provided; (ii) certain types of shortcut reasoning may be missed. To address these issues, we propose a novel method for identifying shortcut reasoning. The proposed method quantifies the severity of the shortcut reasoning by leveraging out-of-distribution data and does not make any assumptions about the type of tokens triggering the shortcut reasoning. Our experiments on Natural Language Inference and Sentiment Analysis demonstrate that our framework successfully discovers known and unknown shortcut reasoning in the previous work.",
}
| Shortcut reasoning is an irrational process of inference, which degrades the robustness of an NLP model. While a number of previous work has tackled the identification of shortcut reasoning, there are still two major limitations: (i) a method for quantifying the severity of the discovered shortcut reasoning is not provided; (ii) certain types of shortcut reasoning may be missed. To address these issues, we propose a novel method for identifying shortcut reasoning. The proposed method quantifies the severity of the shortcut reasoning by leveraging out-of-distribution data and does not make any assumptions about the type of tokens triggering the shortcut reasoning. Our experiments on Natural Language Inference and Sentiment Analysis demonstrate that our framework successfully discovers known and unknown shortcut reasoning in the previous work. | [
"Haraguchi, Daichi",
"Shirai, Kiyoaki",
"Inoue, Naoya",
"Kertkeidkachorn, Natthawut"
] | Discovering Highly Influential Shortcut Reasoning: An Automated Template-Free Approach | findings-emnlp.424 | 2312.09718 | [
"https://github.com/homoscribens/shortcut_reasoning"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.425.bib | https://aclanthology.org/2023.findings-emnlp.425/ | @inproceedings{ye-etal-2023-schema,
title = "Schema-adaptable Knowledge Graph Construction",
author = "Ye, Hongbin and
Gui, Honghao and
Xu, Xin and
Chen, Xi and
Chen, Huajun and
Zhang, Ningyu",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.425",
doi = "10.18653/v1/2023.findings-emnlp.425",
pages = "6408--6431",
abstract = "Conventional Knowledge Graph Construction (KGC) approaches typically follow the static information extraction paradigm with a closed set of pre-defined schema. As a result, such approaches fall short when applied to dynamic scenarios or domains, whereas a new type of knowledge emerges. This necessitates a system that can handle evolving schema automatically to extract information for KGC. To address this need, we propose a new task called schema-adaptable KGC, which aims to continually extract entity, relation, and event based on a dynamically changing schema graph without re-training. We first split and convert existing datasets based on three principles to build a benchmark, i.e., horizontal schema expansion, vertical schema expansion, and hybrid schema expansion; then investigate the schema-adaptable performance of several well-known approaches such as Text2Event, TANL, UIE and GPT-3.5. We further propose a simple yet effective baseline dubbed AdaKGC, which contains schema-enriched prefix instructor and schema-conditioned dynamic decoding to better handle evolving schema. Comprehensive experimental results illustrate that AdaKGC can outperform baselines but still have room for improvement. We hope the proposed work can deliver benefits to the community.",
}
| Conventional Knowledge Graph Construction (KGC) approaches typically follow the static information extraction paradigm with a closed set of pre-defined schema. As a result, such approaches fall short when applied to dynamic scenarios or domains, whereas a new type of knowledge emerges. This necessitates a system that can handle evolving schema automatically to extract information for KGC. To address this need, we propose a new task called schema-adaptable KGC, which aims to continually extract entity, relation, and event based on a dynamically changing schema graph without re-training. We first split and convert existing datasets based on three principles to build a benchmark, i.e., horizontal schema expansion, vertical schema expansion, and hybrid schema expansion; then investigate the schema-adaptable performance of several well-known approaches such as Text2Event, TANL, UIE and GPT-3.5. We further propose a simple yet effective baseline dubbed AdaKGC, which contains schema-enriched prefix instructor and schema-conditioned dynamic decoding to better handle evolving schema. Comprehensive experimental results illustrate that AdaKGC can outperform baselines but still have room for improvement. We hope the proposed work can deliver benefits to the community. | [
"Ye, Hongbin",
"Gui, Honghao",
"Xu, Xin",
"Chen, Xi",
"Chen, Huajun",
"Zhang, Ningyu"
] | Schema-adaptable Knowledge Graph Construction | findings-emnlp.425 | 2305.08703 | [
"https://github.com/zjunlp/adakgc"
] | https://huggingface.co/papers/2305.08703 | 2 | 1 | 0 | 5 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.426.bib | https://aclanthology.org/2023.findings-emnlp.426/ | @inproceedings{veseli-etal-2023-evaluating,
title = "Evaluating the Knowledge Base Completion Potential of {GPT}",
author = "Veseli, Blerta and
Razniewski, Simon and
Kalo, Jan-Christoph and
Weikum, Gerhard",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.426",
doi = "10.18653/v1/2023.findings-emnlp.426",
pages = "6432--6443",
abstract = "Structured knowledge bases (KBs) are an asset for search engines and other applications but are inevitably incomplete. Language models (LMs) have been proposed for unsupervised knowledge base completion (KBC), yet, their ability to do this at scale and with high accuracy remains an open question. Prior experimental studies mostly fall short because they only evaluate on popular subjects, or sample already existing facts from KBs. In this work, we perform a careful evaluation of GPT{'}s potential to complete the largest public KB: Wikidata. We find that, despite their size and capabilities, models like GPT-3, ChatGPT and GPT-4 do not achieve fully convincing results on this task. Nonetheless, it provides solid improvements over earlier approaches with smaller LMs. In particular, we show that it is feasible to extend Wikidata by 27M facts at 90{\%} precision.",
}
| Structured knowledge bases (KBs) are an asset for search engines and other applications but are inevitably incomplete. Language models (LMs) have been proposed for unsupervised knowledge base completion (KBC), yet, their ability to do this at scale and with high accuracy remains an open question. Prior experimental studies mostly fall short because they only evaluate on popular subjects, or sample already existing facts from KBs. In this work, we perform a careful evaluation of GPT{'}s potential to complete the largest public KB: Wikidata. We find that, despite their size and capabilities, models like GPT-3, ChatGPT and GPT-4 do not achieve fully convincing results on this task. Nonetheless, it provides solid improvements over earlier approaches with smaller LMs. In particular, we show that it is feasible to extend Wikidata by 27M facts at 90{\%} precision. | [
"Veseli, Blerta",
"Razniewski, Simon",
"Kalo, Jan-Christoph",
"Weikum, Gerhard"
] | Evaluating the Knowledge Base Completion Potential of GPT | findings-emnlp.426 | 2310.14771 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.427.bib | https://aclanthology.org/2023.findings-emnlp.427/ | @inproceedings{wu-etal-2023-conic10k,
title = "{C}onic10{K}: A Challenging Math Problem Understanding and Reasoning Dataset",
author = "Wu, Haoyi and
Hui, Wenyang and
Chen, Yezeng and
Wu, Weiqi and
Tu, Kewei and
Zhou, Yi",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.427",
doi = "10.18653/v1/2023.findings-emnlp.427",
pages = "6444--6458",
abstract = "Mathematical understanding and reasoning are crucial tasks for assessing the capabilities of artificial intelligence (AI). However, existing benchmarks either require just a few steps of reasoning, or only contain a small amount of data in one specific topic, making it hard to analyse AI{'}s behaviour with reference to different problems within a specific topic in detail. In this work, we propose Conic10K, a challenging math problem dataset on conic sections in Chinese senior high school education. Our dataset contains various problems with different reasoning depths, while only the knowledge from conic sections is required. Since the dataset only involves a narrow range of knowledge, it is easy to separately analyse the knowledge a model possesses and the reasoning ability it has. For each problem, we provide a high-quality formal representation, the reasoning steps, and the final solution. Experiments show that existing large language models, including GPT-4, exhibit weak performance on complex reasoning. We hope that our findings could inspire more advanced techniques for precise natural language understanding and reasoning. Our dataset and codes are available at https://github.com/whyNLP/Conic10K.",
}
| Mathematical understanding and reasoning are crucial tasks for assessing the capabilities of artificial intelligence (AI). However, existing benchmarks either require just a few steps of reasoning, or only contain a small amount of data in one specific topic, making it hard to analyse AI{'}s behaviour with reference to different problems within a specific topic in detail. In this work, we propose Conic10K, a challenging math problem dataset on conic sections in Chinese senior high school education. Our dataset contains various problems with different reasoning depths, while only the knowledge from conic sections is required. Since the dataset only involves a narrow range of knowledge, it is easy to separately analyse the knowledge a model possesses and the reasoning ability it has. For each problem, we provide a high-quality formal representation, the reasoning steps, and the final solution. Experiments show that existing large language models, including GPT-4, exhibit weak performance on complex reasoning. We hope that our findings could inspire more advanced techniques for precise natural language understanding and reasoning. Our dataset and codes are available at https://github.com/whyNLP/Conic10K. | [
"Wu, Haoyi",
"Hui, Wenyang",
"Chen, Yezeng",
"Wu, Weiqi",
"Tu, Kewei",
"Zhou, Yi"
] | Conic10K: A Challenging Math Problem Understanding and Reasoning Dataset | findings-emnlp.427 | 2311.05113 | [
"https://github.com/whynlp/conic10k"
] | https://huggingface.co/papers/2311.05113 | 2 | 1 | 0 | 6 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.428.bib | https://aclanthology.org/2023.findings-emnlp.428/ | @inproceedings{li-etal-2023-depwignn,
title = "{D}ep{W}i{GNN}: A Depth-wise Graph Neural Network for Multi-hop Spatial Reasoning in Text",
author = "Li, Shuaiyi and
Deng, Yang and
Lam, Wai",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.428",
doi = "10.18653/v1/2023.findings-emnlp.428",
pages = "6459--6471",
abstract = "Spatial reasoning in text plays a crucial role in various real-world applications. Existing approaches for spatial reasoning typically infer spatial relations from pure text, which overlook the gap between natural language and symbolic structures. Graph neural networks (GNNs) have showcased exceptional proficiency in inducing and aggregating symbolic structures. However, classical GNNs face challenges in handling multi-hop spatial reasoning due to the over-smoothing issue, i.e., the performance decreases substantially as the number of graph layers increases. To cope with these challenges, we propose a novel Depth-Wise Graph Neural Network (DepWiGNN). Specifically, we design a novel node memory scheme and aggregate the information over the depth dimension instead of the breadth dimension of the graph, which empowers the ability to collect long dependencies without stacking multiple layers. Experimental results on two challenging multi-hop spatial reasoning datasets show that DepWiGNN outperforms existing spatial reasoning methods. The comparisons with the other three GNNs further demonstrate its superiority in capturing long dependency in the graph.",
}
| Spatial reasoning in text plays a crucial role in various real-world applications. Existing approaches for spatial reasoning typically infer spatial relations from pure text, which overlook the gap between natural language and symbolic structures. Graph neural networks (GNNs) have showcased exceptional proficiency in inducing and aggregating symbolic structures. However, classical GNNs face challenges in handling multi-hop spatial reasoning due to the over-smoothing issue, i.e., the performance decreases substantially as the number of graph layers increases. To cope with these challenges, we propose a novel Depth-Wise Graph Neural Network (DepWiGNN). Specifically, we design a novel node memory scheme and aggregate the information over the depth dimension instead of the breadth dimension of the graph, which empowers the ability to collect long dependencies without stacking multiple layers. Experimental results on two challenging multi-hop spatial reasoning datasets show that DepWiGNN outperforms existing spatial reasoning methods. The comparisons with the other three GNNs further demonstrate its superiority in capturing long dependency in the graph. | [
"Li, Shuaiyi",
"Deng, Yang",
"Lam, Wai"
] | DepWiGNN: A Depth-wise Graph Neural Network for Multi-hop Spatial Reasoning in Text | findings-emnlp.428 | 2310.12557 | [
"https://github.com/syon-li/depwignn"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.429.bib | https://aclanthology.org/2023.findings-emnlp.429/ | @inproceedings{botzer-etal-2023-tk,
title = "{TK}-{KNN}: A Balanced Distance-Based Pseudo Labeling Approach for Semi-Supervised Intent Classification",
author = "Botzer, Nicholas and
Vazquez, David and
Weninger, Tim and
Laradji, Issam",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.429",
doi = "10.18653/v1/2023.findings-emnlp.429",
pages = "6472--6484",
abstract = "The ability to detect intent in dialogue systems has become increasingly important in modern technology. These systems often generate a large amount of unlabeled data, and manually labeling this data requires substantial human effort. Semi-supervised methods attempt to remedy this cost by using a model trained on a few labeled examples and then by assigning pseudo-labels to further a subset of unlabeled examples that has a model prediction confidence higher than a certain threshold. However, one particularly perilous consequence of these methods is the risk of picking an imbalanced set of examples across classes, which could lead to poor labels. In the present work, we describe Top-K K-Nearest Neighbor (TK-KNN), which uses a more robust pseudo-labeling approach based on distance in the embedding space while maintaining a balanced set of pseudo-labeled examples across classes through a ranking-based approach. Experiments on several datasets show that TK-KNN outperforms existing models, particularly when labeled data is scarce on popular datasets such as CLINC150 and Banking77.",
}
| The ability to detect intent in dialogue systems has become increasingly important in modern technology. These systems often generate a large amount of unlabeled data, and manually labeling this data requires substantial human effort. Semi-supervised methods attempt to remedy this cost by using a model trained on a few labeled examples and then by assigning pseudo-labels to further a subset of unlabeled examples that has a model prediction confidence higher than a certain threshold. However, one particularly perilous consequence of these methods is the risk of picking an imbalanced set of examples across classes, which could lead to poor labels. In the present work, we describe Top-K K-Nearest Neighbor (TK-KNN), which uses a more robust pseudo-labeling approach based on distance in the embedding space while maintaining a balanced set of pseudo-labeled examples across classes through a ranking-based approach. Experiments on several datasets show that TK-KNN outperforms existing models, particularly when labeled data is scarce on popular datasets such as CLINC150 and Banking77. | [
"Botzer, Nicholas",
"Vazquez, David",
"Weninger, Tim",
"Laradji, Issam"
] | TK-KNN: A Balanced Distance-Based Pseudo Labeling Approach for Semi-Supervised Intent Classification | findings-emnlp.429 | 2310.11607 | [
"https://github.com/servicenow/tk-knn"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.430.bib | https://aclanthology.org/2023.findings-emnlp.430/ | @inproceedings{sharma-etal-2023-late,
title = "Late Fusion of Transformers for Sentiment Analysis of Code-Switched Data",
author = "Sharma, Gagan and
Chinmay, R and
Sharma, Raksha",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.430",
doi = "10.18653/v1/2023.findings-emnlp.430",
pages = "6485--6490",
abstract = "Code-switching is a common phenomenon in multilingual communities and is often used on social media. However, sentiment analysis of code-switched data is a challenging yet less explored area of research. This paper aims to develop a sentiment analysis system for code-switched data. In this paper, we present a novel approach combining two transformers using logits of their output and feeding them to a neural network for classification. We show the efficacy of our approach using two benchmark datasets, viz., English-Hindi (En-Hi), and English-Spanish (En-Es) availed by Microsoft GLUECoS. Our approach results in an F1 score of 73.66{\%} for En-Es and 61.24{\%} for En-Hi, significantly higher than the best model reported for the GLUECoS benchmark dataset.",
}
| Code-switching is a common phenomenon in multilingual communities and is often used on social media. However, sentiment analysis of code-switched data is a challenging yet less explored area of research. This paper aims to develop a sentiment analysis system for code-switched data. In this paper, we present a novel approach combining two transformers using logits of their output and feeding them to a neural network for classification. We show the efficacy of our approach using two benchmark datasets, viz., English-Hindi (En-Hi), and English-Spanish (En-Es) availed by Microsoft GLUECoS. Our approach results in an F1 score of 73.66{\%} for En-Es and 61.24{\%} for En-Hi, significantly higher than the best model reported for the GLUECoS benchmark dataset. | [
"Sharma, Gagan",
"Chinmay, R",
"Sharma, Raksha"
] | Late Fusion of Transformers for Sentiment Analysis of Code-Switched Data | findings-emnlp.430 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.431.bib | https://aclanthology.org/2023.findings-emnlp.431/ | @inproceedings{zhou-etal-2023-inductive,
title = "Inductive Relation Inference of Knowledge Graph Enhanced by Ontology Information",
author = "Zhou, Wentao and
Zhao, Jun and
Gui, Tao and
Zhang, Qi and
Huang, Xuanjing",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.431",
doi = "10.18653/v1/2023.findings-emnlp.431",
pages = "6491--6502",
abstract = "The inductive inference of the knowledge graph aims to complete the potential relations between the new unknown entities in the graph. Most existing methods are based on entity-independent features such as graph structure information and relationship information to inference. However, the neighborhood of these new entities is often too sparse to obtain enough information to build these features effectively. In this work, we propose a knowledge graph inductive inference method that fuses ontology information. Based on the enclosing subgraph, we bring in feature embeddings of concepts corresponding to entities to learn the semantic information implicit in the ontology. Considering that the ontology information of entities may be missing, we build a type constraint regular loss to explicitly model the semantic connections between entities and concepts, and thus capture the missing concepts of entities. Experimental results show that our approach significantly outperforms large language models like ChatGPT on two benchmark datasets, YAGO21K-610 and DB45K-165, and improves the MRR metrics by 15.4{\%} and 44.1{\%}, respectively, when compared with the state-of-the-art methods.",
}
| The inductive inference of the knowledge graph aims to complete the potential relations between the new unknown entities in the graph. Most existing methods are based on entity-independent features such as graph structure information and relationship information to inference. However, the neighborhood of these new entities is often too sparse to obtain enough information to build these features effectively. In this work, we propose a knowledge graph inductive inference method that fuses ontology information. Based on the enclosing subgraph, we bring in feature embeddings of concepts corresponding to entities to learn the semantic information implicit in the ontology. Considering that the ontology information of entities may be missing, we build a type constraint regular loss to explicitly model the semantic connections between entities and concepts, and thus capture the missing concepts of entities. Experimental results show that our approach significantly outperforms large language models like ChatGPT on two benchmark datasets, YAGO21K-610 and DB45K-165, and improves the MRR metrics by 15.4{\%} and 44.1{\%}, respectively, when compared with the state-of-the-art methods. | [
"Zhou, Wentao",
"Zhao, Jun",
"Gui, Tao",
"Zhang, Qi",
"Huang, Xuanjing"
] | Inductive Relation Inference of Knowledge Graph Enhanced by Ontology Information | findings-emnlp.431 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.432.bib | https://aclanthology.org/2023.findings-emnlp.432/ | @inproceedings{figueras-etal-2023-dynamic,
title = "Dynamic Stance: Modeling Discussions by Labeling the Interactions",
author = "Figueras, Blanca and
Baucells, Irene and
Caselli, Tommaso",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.432",
doi = "10.18653/v1/2023.findings-emnlp.432",
pages = "6503--6515",
abstract = "Stance detection is an increasingly popular task that has been mainly modeled as a static task, by assigning the expressed attitude of a text toward a given topic. Such a framing presents limitations, with trained systems showing poor generalization capabilities and being strongly topic-dependent. In this work, we propose modeling stance as a dynamic task, by focusing on the interactions between a message and their replies. For this purpose, we present a new annotation scheme that enables the categorization of all kinds of textual interactions. As a result, we have created a new corpus, the Dynamic Stance Corpus (DySC), consisting of three datasets in two middle-resourced languages: Catalan and Dutch. Our data analysis further supports our modeling decisions, empirically showing differences between the annotation of stance in static and dynamic contexts. We fine-tuned a series of monolingual and multilingual models on DySC, showing portability across topics and languages.",
}
| Stance detection is an increasingly popular task that has been mainly modeled as a static task, by assigning the expressed attitude of a text toward a given topic. Such a framing presents limitations, with trained systems showing poor generalization capabilities and being strongly topic-dependent. In this work, we propose modeling stance as a dynamic task, by focusing on the interactions between a message and their replies. For this purpose, we present a new annotation scheme that enables the categorization of all kinds of textual interactions. As a result, we have created a new corpus, the Dynamic Stance Corpus (DySC), consisting of three datasets in two middle-resourced languages: Catalan and Dutch. Our data analysis further supports our modeling decisions, empirically showing differences between the annotation of stance in static and dynamic contexts. We fine-tuned a series of monolingual and multilingual models on DySC, showing portability across topics and languages. | [
"Figueras, Blanca",
"Baucells, Irene",
"Caselli, Tommaso"
] | Dynamic Stance: Modeling Discussions by Labeling the Interactions | findings-emnlp.432 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.433.bib | https://aclanthology.org/2023.findings-emnlp.433/ | @inproceedings{qian-etal-2023-harnessing,
title = "Harnessing the Power of Large Language Models for Empathetic Response Generation: Empirical Investigations and Improvements",
author = "Qian, Yushan and
Zhang, Weinan and
Liu, Ting",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.433",
doi = "10.18653/v1/2023.findings-emnlp.433",
pages = "6516--6528",
abstract = "Empathetic dialogue is an indispensable part of building harmonious social relationships and contributes to the development of a helpful AI. Previous approaches are mainly based on fine small-scale language models. With the advent of ChatGPT, the application effect of large language models (LLMs) in this field has attracted great attention. This work empirically investigates the performance of LLMs in generating empathetic responses and proposes three improvement methods of semantically similar in-context learning, two-stage interactive generation, and combination with the knowledge base. Extensive experiments show that LLMs can significantly benefit from our proposed methods and is able to achieve state-of-the-art performance in both automatic and human evaluations. Additionally, we explore the possibility of GPT-4 simulating human evaluators.",
}
| Empathetic dialogue is an indispensable part of building harmonious social relationships and contributes to the development of a helpful AI. Previous approaches are mainly based on fine small-scale language models. With the advent of ChatGPT, the application effect of large language models (LLMs) in this field has attracted great attention. This work empirically investigates the performance of LLMs in generating empathetic responses and proposes three improvement methods of semantically similar in-context learning, two-stage interactive generation, and combination with the knowledge base. Extensive experiments show that LLMs can significantly benefit from our proposed methods and is able to achieve state-of-the-art performance in both automatic and human evaluations. Additionally, we explore the possibility of GPT-4 simulating human evaluators. | [
"Qian, Yushan",
"Zhang, Weinan",
"Liu, Ting"
] | Harnessing the Power of Large Language Models for Empathetic Response Generation: Empirical Investigations and Improvements | findings-emnlp.433 | 2310.05140 | [
"https://github.com/27182812/LLM4ED"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.434.bib | https://aclanthology.org/2023.findings-emnlp.434/ | @inproceedings{peskoff-etal-2023-gpt,
title = "{GPT} Deciphering Fedspeak: Quantifying Dissent Among Hawks and Doves",
author = "Peskoff, Denis and
Visokay, Adam and
Schulhoff, Sander and
Wachspress, Benjamin and
Blinder, Alan and
Stewart, Brandon",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.434",
doi = "10.18653/v1/2023.findings-emnlp.434",
pages = "6529--6539",
abstract = "Markets and policymakers around the world hang on the consequential monetary policy decisions made by the Federal Open Market Committee (FOMC). Publicly available textual documentation of their meetings provides insight into members{'} attitudes about the economy. We use GPT-4 to quantify dissent among members on the topic of inflation. We find that transcripts and minutes reflect the diversity of member views about the macroeconomic outlook in a way that is lost or omitted from the public statements. In fact, diverging opinions that shed light upon the committee{'}s {``}true{''} attitudes are almost entirely omitted from the final statements. Hence, we argue that forecasting FOMC sentiment based solely on statements will not sufficiently reflect dissent among the hawks and doves.",
}
| Markets and policymakers around the world hang on the consequential monetary policy decisions made by the Federal Open Market Committee (FOMC). Publicly available textual documentation of their meetings provides insight into members{'} attitudes about the economy. We use GPT-4 to quantify dissent among members on the topic of inflation. We find that transcripts and minutes reflect the diversity of member views about the macroeconomic outlook in a way that is lost or omitted from the public statements. In fact, diverging opinions that shed light upon the committee{'}s {``}true{''} attitudes are almost entirely omitted from the final statements. Hence, we argue that forecasting FOMC sentiment based solely on statements will not sufficiently reflect dissent among the hawks and doves. | [
"Peskoff, Denis",
"Visokay, Adam",
"Schulhoff, S",
"er",
"Wachspress, Benjamin",
"Blinder, Alan",
"Stewart, Br",
"on"
] | GPT Deciphering Fedspeak: Quantifying Dissent Among Hawks and Doves | findings-emnlp.434 | 2407.19110 | [
"https://github.com/DenisPeskoff/FedNLP"
] | https://huggingface.co/papers/2407.19110 | 1 | 1 | 0 | 6 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.435.bib | https://aclanthology.org/2023.findings-emnlp.435/ | @inproceedings{zheng-etal-2023-dialogqae,
title = "{D}ialog{QAE}: N-to-N Question Answer Pair Extraction from Customer Service Chatlog",
author = "Zheng, Xin and
Liu, Tianyu and
Meng, Haoran and
Wang, Xu and
Jiang, Yufan and
Rao, Mengliang and
Lin, Binghuai and
Cao, Yunbo and
Sui, Zhifang",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.435",
doi = "10.18653/v1/2023.findings-emnlp.435",
pages = "6540--6558",
abstract = "Harvesting question-answer (QA) pairs from customer service chatlog in the wild is an efficient way to enrich the knowledge base for customer service chatbots in the cold start or continuous integration scenarios. Prior work attempts to obtain 1-to-1 QA pairs from growing customer service chatlog, which fails to integrate the incomplete utterances from the dialog context for composite QA retrieval. In this paper, we propose N-to-N QA extraction task in which the derived questions and corresponding answers might be separated across different utterances. We introduce a suite of generative/discriminative tagging based methods with end-to-end and two-stage variants that perform well on 5 customer service datasets and for the first time setup a benchmark for N-to-N DialogQAE with utterance and session level evaluation metrics. With a deep dive into extracted QA pairs, we find that the relations between and inside the QA pairs can be indicators to analyze the dialogue structure, e.g. information seeking, clarification, barge-in and elaboration. We also show that the proposed models can adapt to different domains and languages, and reduce the labor cost of knowledge accumulation in the real-world product dialogue platform.",
}
| Harvesting question-answer (QA) pairs from customer service chatlog in the wild is an efficient way to enrich the knowledge base for customer service chatbots in the cold start or continuous integration scenarios. Prior work attempts to obtain 1-to-1 QA pairs from growing customer service chatlog, which fails to integrate the incomplete utterances from the dialog context for composite QA retrieval. In this paper, we propose N-to-N QA extraction task in which the derived questions and corresponding answers might be separated across different utterances. We introduce a suite of generative/discriminative tagging based methods with end-to-end and two-stage variants that perform well on 5 customer service datasets and for the first time setup a benchmark for N-to-N DialogQAE with utterance and session level evaluation metrics. With a deep dive into extracted QA pairs, we find that the relations between and inside the QA pairs can be indicators to analyze the dialogue structure, e.g. information seeking, clarification, barge-in and elaboration. We also show that the proposed models can adapt to different domains and languages, and reduce the labor cost of knowledge accumulation in the real-world product dialogue platform. | [
"Zheng, Xin",
"Liu, Tianyu",
"Meng, Haoran",
"Wang, Xu",
"Jiang, Yufan",
"Rao, Mengliang",
"Lin, Binghuai",
"Cao, Yunbo",
"Sui, Zhifang"
] | DialogQAE: N-to-N Question Answer Pair Extraction from Customer Service Chatlog | findings-emnlp.435 | 2212.07112 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.436.bib | https://aclanthology.org/2023.findings-emnlp.436/ | @inproceedings{fu-etal-2023-inverse,
title = "Inverse Reinforcement Learning for Text Summarization",
author = "Fu, Yu and
Xiong, Deyi and
Dong, Yue",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.436",
doi = "10.18653/v1/2023.findings-emnlp.436",
pages = "6559--6570",
abstract = "We introduce inverse reinforcement learning (IRL) as an effective paradigm for training abstractive summarization models, imitating human summarization behaviors. Our IRL model estimates the reward function using a suite of important sub-rewards for summarization and concurrently optimizes the policy network. Experimental results across datasets in different domains (CNN/DailyMail and WikiHow) and various model sizes (BART-base and BART-large) demonstrate the superiority of our proposed IRL model for summarization over MLE and RL baselines. The resulting summaries exhibit greater similarity to human-crafted gold references, outperforming MLE and RL baselines on metrics such as ROUGE, coverage, novelty, compression ratio, factuality, and human evaluations.",
}
| We introduce inverse reinforcement learning (IRL) as an effective paradigm for training abstractive summarization models, imitating human summarization behaviors. Our IRL model estimates the reward function using a suite of important sub-rewards for summarization and concurrently optimizes the policy network. Experimental results across datasets in different domains (CNN/DailyMail and WikiHow) and various model sizes (BART-base and BART-large) demonstrate the superiority of our proposed IRL model for summarization over MLE and RL baselines. The resulting summaries exhibit greater similarity to human-crafted gold references, outperforming MLE and RL baselines on metrics such as ROUGE, coverage, novelty, compression ratio, factuality, and human evaluations. | [
"Fu, Yu",
"Xiong, Deyi",
"Dong, Yue"
] | Inverse Reinforcement Learning for Text Summarization | findings-emnlp.436 | 2212.09917 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.437.bib | https://aclanthology.org/2023.findings-emnlp.437/ | @inproceedings{khademi-etal-2023-mm,
title = "{MM}-Reasoner: A Multi-Modal Knowledge-Aware Framework for Knowledge-Based Visual Question Answering",
author = "Khademi, Mahmoud and
Yang, Ziyi and
Frujeri, Felipe and
Zhu, Chenguang",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.437",
doi = "10.18653/v1/2023.findings-emnlp.437",
pages = "6571--6581",
abstract = "Thanks to the strong reasoning capabilities of Large Language Models (LLMs), recent approaches to knowledge-based visual question answering (KVQA) utilize LLMs with a global caption of an input image to answer a question. However, these approaches may miss key visual information that is not captured by the caption. Moreover, they cannot fully utilize the visual information required to answer the question. To address these issues, we introduce a new framework called Multi-Modal Knowledge-Aware Reasoner (MM-Reasoner) for KVQA. MM-Reasoner first utilizes a set of vision APIs, such as dense captioners, object detectors, and OCR, to extract detailed information from the image in textual format. Then, it prompts an LLM to extract query-specific knowledge from the extracted textual information to provide a rich representation that contains external knowledge, commonsense, explicit supporting facts, and rationales required for reasoning. Finally, the knowledge, query, and visual input are used to fine-tune a Vision-Language Model (VLM). At test time, MM-Reasoner uses the potential answers predicted by the VLM to iteratively update and optimize the prompt, refining its answer. Empirical studies show that MM-Reasoner achieves state-of-the-art performance on several KVQA datasets.",
}
| Thanks to the strong reasoning capabilities of Large Language Models (LLMs), recent approaches to knowledge-based visual question answering (KVQA) utilize LLMs with a global caption of an input image to answer a question. However, these approaches may miss key visual information that is not captured by the caption. Moreover, they cannot fully utilize the visual information required to answer the question. To address these issues, we introduce a new framework called Multi-Modal Knowledge-Aware Reasoner (MM-Reasoner) for KVQA. MM-Reasoner first utilizes a set of vision APIs, such as dense captioners, object detectors, and OCR, to extract detailed information from the image in textual format. Then, it prompts an LLM to extract query-specific knowledge from the extracted textual information to provide a rich representation that contains external knowledge, commonsense, explicit supporting facts, and rationales required for reasoning. Finally, the knowledge, query, and visual input are used to fine-tune a Vision-Language Model (VLM). At test time, MM-Reasoner uses the potential answers predicted by the VLM to iteratively update and optimize the prompt, refining its answer. Empirical studies show that MM-Reasoner achieves state-of-the-art performance on several KVQA datasets. | [
"Khademi, Mahmoud",
"Yang, Ziyi",
"Frujeri, Felipe",
"Zhu, Chenguang"
] | MM-Reasoner: A Multi-Modal Knowledge-Aware Framework for Knowledge-Based Visual Question Answering | findings-emnlp.437 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.438.bib | https://aclanthology.org/2023.findings-emnlp.438/ | @inproceedings{chou-etal-2023-toward,
title = "Toward Joint Language Modeling for Speech Units and Text",
author = "Chou, Ju-Chieh and
Chien, Chung-Ming and
Hsu, Wei-Ning and
Livescu, Karen and
Babu, Arun and
Conneau, Alexis and
Baevski, Alexei and
Auli, Michael",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.438",
doi = "10.18653/v1/2023.findings-emnlp.438",
pages = "6582--6593",
abstract = "Speech and text are two major forms of human language. The research community has been focusing on mapping speech to text or vice versa for many years. However, in the field of language modeling, very little effort has been made to model them jointly. In light of this, we explore joint language modeling for speech units and text. Specifically, we compare different speech tokenizers to transform continuous speech signals into discrete units and use different methods to construct mixed speech-text data. We introduce automatic metrics to evaluate how well the joint LM mixes speech and text. We also fine-tune the LM on downstream spoken language understanding (SLU) tasks with different modalities (speech or text) and test its performance to assess the model{'}s learning of shared representations. Our results show that by mixing speech units and text with our proposed mixing techniques, the joint LM improves over a speech-only baseline on SLU tasks and shows zero-shot cross-modal transferability.",
}
| Speech and text are two major forms of human language. The research community has been focusing on mapping speech to text or vice versa for many years. However, in the field of language modeling, very little effort has been made to model them jointly. In light of this, we explore joint language modeling for speech units and text. Specifically, we compare different speech tokenizers to transform continuous speech signals into discrete units and use different methods to construct mixed speech-text data. We introduce automatic metrics to evaluate how well the joint LM mixes speech and text. We also fine-tune the LM on downstream spoken language understanding (SLU) tasks with different modalities (speech or text) and test its performance to assess the model{'}s learning of shared representations. Our results show that by mixing speech units and text with our proposed mixing techniques, the joint LM improves over a speech-only baseline on SLU tasks and shows zero-shot cross-modal transferability. | [
"Chou, Ju-Chieh",
"Chien, Chung-Ming",
"Hsu, Wei-Ning",
"Livescu, Karen",
"Babu, Arun",
"Conneau, Alexis",
"Baevski, Alexei",
"Auli, Michael"
] | Toward Joint Language Modeling for Speech Units and Text | findings-emnlp.438 | 2310.08715 | [
""
] | https://huggingface.co/papers/2310.08715 | 6 | 7 | 1 | 8 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.439.bib | https://aclanthology.org/2023.findings-emnlp.439/ | @inproceedings{sundriyal-etal-2023-chaos,
title = "From Chaos to Clarity: Claim Normalization to Empower Fact-Checking",
author = "Sundriyal, Megha and
Chakraborty, Tanmoy and
Nakov, Preslav",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.439",
doi = "10.18653/v1/2023.findings-emnlp.439",
pages = "6594--6609",
abstract = "With the proliferation of social media platforms, users are exposed to vast information, including posts containing misleading claims. However, the pervasive noise inherent in these posts presents a challenge in identifying precise and prominent claims that require verification. Extracting the core assertions from such posts is arduous and time-consuming. We introduce a novel task, called Claim Normalization ($\textit{aka ClaimNorm}$) that aims to decompose complex and noisy social media posts into more straightforward and understandable forms, termed $\textit{normalized claims}$. We propose $\texttt{CACN}$ , a pioneering approach that leverages chain-of-thought and claim check-worthiness estimation, mimicking human reasoning processes, to comprehend intricate claims. Moreover, we capitalize on large language models{'} powerful in-context learning abilities to provide guidance and improve the claim normalization process. To evaluate the effectiveness of our proposed model, we meticulously compile a comprehensive real-world dataset, $\texttt{CLAN}$, comprising more than $6k$ instances of social media posts alongside their respective normalized claims. Experimentation demonstrates that $\texttt{CACN}$ outperforms several baselines across various evaluation measures. A rigorous error analysis validates $\texttt{CACN}${`}s capabilities and pitfalls. We release our dataset and code at https://github.com/LCS2-IIITD/CACN-EMNLP-2023.",
}
| With the proliferation of social media platforms, users are exposed to vast information, including posts containing misleading claims. However, the pervasive noise inherent in these posts presents a challenge in identifying precise and prominent claims that require verification. Extracting the core assertions from such posts is arduous and time-consuming. We introduce a novel task, called Claim Normalization ($\textit{aka ClaimNorm}$) that aims to decompose complex and noisy social media posts into more straightforward and understandable forms, termed $\textit{normalized claims}$. We propose $\texttt{CACN}$ , a pioneering approach that leverages chain-of-thought and claim check-worthiness estimation, mimicking human reasoning processes, to comprehend intricate claims. Moreover, we capitalize on large language models{'} powerful in-context learning abilities to provide guidance and improve the claim normalization process. To evaluate the effectiveness of our proposed model, we meticulously compile a comprehensive real-world dataset, $\texttt{CLAN}$, comprising more than $6k$ instances of social media posts alongside their respective normalized claims. Experimentation demonstrates that $\texttt{CACN}$ outperforms several baselines across various evaluation measures. A rigorous error analysis validates $\texttt{CACN}${`}s capabilities and pitfalls. We release our dataset and code at https://github.com/LCS2-IIITD/CACN-EMNLP-2023. | [
"Sundriyal, Megha",
"Chakraborty, Tanmoy",
"Nakov, Preslav"
] | From Chaos to Clarity: Claim Normalization to Empower Fact-Checking | findings-emnlp.439 | 2310.14338 | [
"https://github.com/lcs2-iiitd/cacn-emnlp-2023"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.440.bib | https://aclanthology.org/2023.findings-emnlp.440/ | @inproceedings{zhang-etal-2023-mitigating,
title = "Mitigating Biases in Hate Speech Detection from A Causal Perspective",
author = "Zhang, Zhehao and
Chen, Jiaao and
Yang, Diyi",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.440",
doi = "10.18653/v1/2023.findings-emnlp.440",
pages = "6610--6625",
abstract = "Nowadays, many hate speech detectors are built to automatically detect hateful content. However, their training sets are sometimes skewed towards certain stereotypes (e.g., race or religion-related). As a result, the detectors are prone to depend on some shortcuts for predictions. Previous works mainly focus on token-level analysis and heavily rely on human experts{'} annotations to identify spurious correlations, which is not only costly but also incapable of discovering higher-level artifacts. In this work, we use grammar induction to find grammar patterns for hate speech and analyze this phenomenon from a causal perspective. Concretely, we categorize and verify different biases based on their spuriousness and influence on the model prediction. Then, we propose two mitigation approaches including Multi-Task Intervention and Data-Specific Intervention based on these confounders. Experiments conducted on 9 hate speech datasets demonstrate the effectiveness of our approaches.",
}
| Nowadays, many hate speech detectors are built to automatically detect hateful content. However, their training sets are sometimes skewed towards certain stereotypes (e.g., race or religion-related). As a result, the detectors are prone to depend on some shortcuts for predictions. Previous works mainly focus on token-level analysis and heavily rely on human experts{'} annotations to identify spurious correlations, which is not only costly but also incapable of discovering higher-level artifacts. In this work, we use grammar induction to find grammar patterns for hate speech and analyze this phenomenon from a causal perspective. Concretely, we categorize and verify different biases based on their spuriousness and influence on the model prediction. Then, we propose two mitigation approaches including Multi-Task Intervention and Data-Specific Intervention based on these confounders. Experiments conducted on 9 hate speech datasets demonstrate the effectiveness of our approaches. | [
"Zhang, Zhehao",
"Chen, Jiaao",
"Yang, Diyi"
] | Mitigating Biases in Hate Speech Detection from A Causal Perspective | findings-emnlp.440 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.441.bib | https://aclanthology.org/2023.findings-emnlp.441/ | @inproceedings{ocampo-etal-2023-unmasking,
title = "Unmasking the Hidden Meaning: Bridging Implicit and Explicit Hate Speech Embedding Representations",
author = "Ocampo, Nicol{\'a}s Benjam{\'\i}n and
Cabrio, Elena and
Villata, Serena",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.441",
doi = "10.18653/v1/2023.findings-emnlp.441",
pages = "6626--6637",
abstract = "Research on automatic hate speech (HS) detection has mainly focused on identifying explicit forms of hateful expressions on user-generated content. Recently, a few works have started to investigate methods to address more implicit and subtle abusive content. However, despite these efforts, automated systems still struggle to correctly recognize implicit and more veiled forms of HS. As these systems heavily rely on proper textual representations for classification, it is crucial to investigate the differences in embedding implicit and explicit messages. Our contribution to address this challenging task is fourfold. First, we present a comparative analysis of transformer-based models, evaluating their performance across five datasets containing implicit HS messages. Second, we examine the embedding representations of implicit messages across different targets, gaining insight into how veiled cases are encoded. Third, we compare and link explicit and implicit hateful messages across these datasets through their targets, enforcing the relation between explicitness and implicitness and obtaining more meaningful embedding representations. Lastly, we show how these newer representation maintains high performance on HS labels, while improving classification in borderline cases.",
}
| Research on automatic hate speech (HS) detection has mainly focused on identifying explicit forms of hateful expressions on user-generated content. Recently, a few works have started to investigate methods to address more implicit and subtle abusive content. However, despite these efforts, automated systems still struggle to correctly recognize implicit and more veiled forms of HS. As these systems heavily rely on proper textual representations for classification, it is crucial to investigate the differences in embedding implicit and explicit messages. Our contribution to address this challenging task is fourfold. First, we present a comparative analysis of transformer-based models, evaluating their performance across five datasets containing implicit HS messages. Second, we examine the embedding representations of implicit messages across different targets, gaining insight into how veiled cases are encoded. Third, we compare and link explicit and implicit hateful messages across these datasets through their targets, enforcing the relation between explicitness and implicitness and obtaining more meaningful embedding representations. Lastly, we show how these newer representation maintains high performance on HS labels, while improving classification in borderline cases. | [
"Ocampo, Nicol{\\'a}s Benjam{\\'\\i}n",
"Cabrio, Elena",
"Villata, Serena"
] | Unmasking the Hidden Meaning: Bridging Implicit and Explicit Hate Speech Embedding Representations | findings-emnlp.441 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.442.bib | https://aclanthology.org/2023.findings-emnlp.442/ | @inproceedings{li-etal-2023-perturbscore,
title = "{P}erturb{S}core: Connecting Discrete and Continuous Perturbations in {NLP}",
author = "Li, Linyang and
Ren, Ke and
Shao, Yunfan and
Wang, Pengyu and
Qiu, Xipeng",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.442",
doi = "10.18653/v1/2023.findings-emnlp.442",
pages = "6638--6648",
abstract = "With the rapid development of neural network applications in NLP, model robustness problem is gaining more attention. Different from computer vision, the discrete nature of texts makes it more challenging to explore robustness in NLP. Therefore, in this paper, we aim to connect discrete perturbations with continuous perturbations, therefore we can use such connections as a bridge to help understand discrete perturbations in NLP models. Specifically, we first explore how to connect and measure the correlation between discrete perturbations and continuous perturbations. Then we design a regression task as a PerturbScore to learn the correlation automatically. Through experimental results, we find that we can build a connection between discrete and continuous perturbations and use the proposed PerturbScore to learn such correlation, surpassing previous methods used in discrete perturbation measuring. Further, the proposed PerturbScore can be well generalized to different datasets, perturbation methods, indicating that we can use it as a powerful tool to study model robustness in NLP.",
}
| With the rapid development of neural network applications in NLP, model robustness problem is gaining more attention. Different from computer vision, the discrete nature of texts makes it more challenging to explore robustness in NLP. Therefore, in this paper, we aim to connect discrete perturbations with continuous perturbations, therefore we can use such connections as a bridge to help understand discrete perturbations in NLP models. Specifically, we first explore how to connect and measure the correlation between discrete perturbations and continuous perturbations. Then we design a regression task as a PerturbScore to learn the correlation automatically. Through experimental results, we find that we can build a connection between discrete and continuous perturbations and use the proposed PerturbScore to learn such correlation, surpassing previous methods used in discrete perturbation measuring. Further, the proposed PerturbScore can be well generalized to different datasets, perturbation methods, indicating that we can use it as a powerful tool to study model robustness in NLP. | [
"Li, Linyang",
"Ren, Ke",
"Shao, Yunfan",
"Wang, Pengyu",
"Qiu, Xipeng"
] | PerturbScore: Connecting Discrete and Continuous Perturbations in NLP | findings-emnlp.442 | 2310.08889 | [
"https://github.com/renke999/perturbscore"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.443.bib | https://aclanthology.org/2023.findings-emnlp.443/ | @inproceedings{jin-etal-2023-instructor,
title = "{I}nstructo{R}: Instructing Unsupervised Conversational Dense Retrieval with Large Language Models",
author = "Jin, Zhuoran and
Cao, Pengfei and
Chen, Yubo and
Liu, Kang and
Zhao, Jun",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.443",
doi = "10.18653/v1/2023.findings-emnlp.443",
pages = "6649--6675",
abstract = "Compared to traditional single-turn ad-hoc retrieval, conversational retrieval needs to handle the multi-turn conversation and understand the user{'}s real query intent. However, most existing methods simply fine-tune the pre-trained ad-hoc retriever on limited supervised data, making it challenging for the retriever to fully grasp the entirety of the conversation. In this paper, we find that large language models (LLMs) can accurately discover the user{'}s query intent from the complex conversation context and provide the supervised signal to instruct the retriever in an unsupervised manner. Therefore, we propose a novel method termed InstructoR to Instruct unsupervised conversational dense Retrieval with LLMs. We design an unsupervised training framework that employs LLMs to estimate the session-passage relevance score as the soft label to guide the retriever{'}s training. Specially, we devise three instructing strategies from context, query and response perspectives to calculate the relevance score more precisely, including conversational retrieval as conversation generation, question rewrite as latent variable and question response as posterior guide. Experimental results show InstructoR can bring significant improvements across various ad-hoc retrievers, even surpassing the current supervised state-of-the-art method. We also demonstrate the effectiveness of our method under low-resource and zero-shot settings. Our code is publicly available at https://github.com/jinzhuoran/InstructoR/.",
}
| Compared to traditional single-turn ad-hoc retrieval, conversational retrieval needs to handle the multi-turn conversation and understand the user{'}s real query intent. However, most existing methods simply fine-tune the pre-trained ad-hoc retriever on limited supervised data, making it challenging for the retriever to fully grasp the entirety of the conversation. In this paper, we find that large language models (LLMs) can accurately discover the user{'}s query intent from the complex conversation context and provide the supervised signal to instruct the retriever in an unsupervised manner. Therefore, we propose a novel method termed InstructoR to Instruct unsupervised conversational dense Retrieval with LLMs. We design an unsupervised training framework that employs LLMs to estimate the session-passage relevance score as the soft label to guide the retriever{'}s training. Specially, we devise three instructing strategies from context, query and response perspectives to calculate the relevance score more precisely, including conversational retrieval as conversation generation, question rewrite as latent variable and question response as posterior guide. Experimental results show InstructoR can bring significant improvements across various ad-hoc retrievers, even surpassing the current supervised state-of-the-art method. We also demonstrate the effectiveness of our method under low-resource and zero-shot settings. Our code is publicly available at https://github.com/jinzhuoran/InstructoR/. | [
"Jin, Zhuoran",
"Cao, Pengfei",
"Chen, Yubo",
"Liu, Kang",
"Zhao, Jun"
] | InstructoR: Instructing Unsupervised Conversational Dense Retrieval with Large Language Models | findings-emnlp.443 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.444.bib | https://aclanthology.org/2023.findings-emnlp.444/ | @inproceedings{loakman-etal-2023-iron,
title = "The Iron(ic) Melting Pot: Reviewing Human Evaluation in Humour, Irony and Sarcasm Generation",
author = "Loakman, Tyler and
Maladry, Aaron and
Lin, Chenghua",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.444",
doi = "10.18653/v1/2023.findings-emnlp.444",
pages = "6676--6689",
abstract = "Human evaluation in often considered to be the gold standard method of evaluating a Natural Language Generation system. However, whilst its importance is accepted by the community at large, the quality of its execution is often brought into question. In this position paper, we argue that the generation of more esoteric forms of language - humour, irony and sarcasm - constitutes a subdomain where the characteristics of selected evaluator panels are of utmost importance, and every effort should be made to report demographic characteristics wherever possible, in the interest of transparency and replicability. We support these claims with an overview of each language form and an analysis of examples in terms of how their interpretation is affected by different participant variables. We additionally perform a critical survey of recent works in NLG to assess how well evaluation procedures are reported in this subdomain, and note a severe lack of open reporting of evaluator demographic information, and a significant reliance on crowdsourcing platforms for recruitment.",
}
| Human evaluation in often considered to be the gold standard method of evaluating a Natural Language Generation system. However, whilst its importance is accepted by the community at large, the quality of its execution is often brought into question. In this position paper, we argue that the generation of more esoteric forms of language - humour, irony and sarcasm - constitutes a subdomain where the characteristics of selected evaluator panels are of utmost importance, and every effort should be made to report demographic characteristics wherever possible, in the interest of transparency and replicability. We support these claims with an overview of each language form and an analysis of examples in terms of how their interpretation is affected by different participant variables. We additionally perform a critical survey of recent works in NLG to assess how well evaluation procedures are reported in this subdomain, and note a severe lack of open reporting of evaluator demographic information, and a significant reliance on crowdsourcing platforms for recruitment. | [
"Loakman, Tyler",
"Maladry, Aaron",
"Lin, Chenghua"
] | The Iron(ic) Melting Pot: Reviewing Human Evaluation in Humour, Irony and Sarcasm Generation | findings-emnlp.444 | 2311.05552 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.445.bib | https://aclanthology.org/2023.findings-emnlp.445/ | @inproceedings{renduchintala-etal-2023-ingenious,
title = "{INGENIOUS}: Using Informative Data Subsets for Efficient Pre-Training of Language Models",
author = "Renduchintala, H S V N S Kowndinya and
Killamsetty, Krishnateja and
Bhatia, Sumit and
Aggarwal, Milan and
Ramakrishnan, Ganesh and
Iyer, Rishabh and
Krishnamurthy, Balaji",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.445",
doi = "10.18653/v1/2023.findings-emnlp.445",
pages = "6690--6705",
abstract = "A salient characteristic of pre-trained language models (PTLMs) is a remarkable improvement in their generalization capability and emergence of new capabilities with increasing model capacity and pre-training dataset size. Consequently, we are witnessing the development of enormous models pushing the state-of-the-art. It is, however, imperative to realize that this inevitably leads to prohibitively long training times, extortionate computing costs, and a detrimental environmental impact. Significant efforts are underway to make PTLM training more efficient through innovations in model architectures, training pipelines, and loss function design, with scant attention being paid to optimizing the utility of training data. The key question that we ask is whether it is possible to train PTLMs by employing only highly informative subsets of the training data while maintaining downstream performance? Building upon the recent progress in informative data subset selection, we show how we can employ submodular optimization to select highly representative subsets of the training corpora and demonstrate that the proposed framework can be applied to efficiently train multiple PTLMs (BERT, BioBERT, GPT-2) using only a fraction of data. Further, we perform a rigorous empirical evaluation to show that the resulting models achieve up to {\textasciitilde}99{\%} of the performance of the fully-trained models. We made our framework publicly available at \url{https://github.com/Efficient-AI/ingenious}.",
}
| A salient characteristic of pre-trained language models (PTLMs) is a remarkable improvement in their generalization capability and emergence of new capabilities with increasing model capacity and pre-training dataset size. Consequently, we are witnessing the development of enormous models pushing the state-of-the-art. It is, however, imperative to realize that this inevitably leads to prohibitively long training times, extortionate computing costs, and a detrimental environmental impact. Significant efforts are underway to make PTLM training more efficient through innovations in model architectures, training pipelines, and loss function design, with scant attention being paid to optimizing the utility of training data. The key question that we ask is whether it is possible to train PTLMs by employing only highly informative subsets of the training data while maintaining downstream performance? Building upon the recent progress in informative data subset selection, we show how we can employ submodular optimization to select highly representative subsets of the training corpora and demonstrate that the proposed framework can be applied to efficiently train multiple PTLMs (BERT, BioBERT, GPT-2) using only a fraction of data. Further, we perform a rigorous empirical evaluation to show that the resulting models achieve up to {\textasciitilde}99{\%} of the performance of the fully-trained models. We made our framework publicly available at \url{https://github.com/Efficient-AI/ingenious}. | [
"Renduchintala, H S V N S Kowndinya",
"Killamsetty, Krishnateja",
"Bhatia, Sumit",
"Aggarwal, Milan",
"Ramakrishnan, Ganesh",
"Iyer, Rishabh",
"Krishnamurthy, Balaji"
] | INGENIOUS: Using Informative Data Subsets for Efficient Pre-Training of Language Models | findings-emnlp.445 | 2305.06677 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.446.bib | https://aclanthology.org/2023.findings-emnlp.446/ | @inproceedings{wu-etal-2023-towards-general,
title = "Towards General Error Diagnosis via Behavioral Testing in Machine Translation",
author = "Wu, Junjie and
Liu, Lemao and
Yeung, Dit-Yan",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.446",
doi = "10.18653/v1/2023.findings-emnlp.446",
pages = "6706--6720",
abstract = "Behavioral testing offers a crucial means of diagnosing linguistic errors and assessing capabilities of NLP models. However, applying behavioral testing to machine translation (MT) systems is challenging as it generally requires human efforts to craft references for evaluating the translation quality of such systems on newly generated test cases. Existing works in behavioral testing of MT systems circumvent this by evaluating translation quality without references, but this restricts diagnosis to specific types of errors, such as incorrect translation of single numeric or currency words. In order to diagnose general errors, this paper proposes a new Bilingual Translation Pair Generation based Behavior Testing (BTPGBT) framework for conducting behavioral testing of MT systems. The core idea of BTPGBT is to employ a novel bilingual translation pair generation (BTPG) approach that automates the construction of high-quality test cases and their pseudoreferences. Experimental results on various MT systems demonstrate that BTPGBT could provide comprehensive and accurate behavioral testing results for general error diagnosis, which further leads to several insightful findings. Our code and data are available at https: //github.com/wujunjie1998/BTPGBT.",
}
| Behavioral testing offers a crucial means of diagnosing linguistic errors and assessing capabilities of NLP models. However, applying behavioral testing to machine translation (MT) systems is challenging as it generally requires human efforts to craft references for evaluating the translation quality of such systems on newly generated test cases. Existing works in behavioral testing of MT systems circumvent this by evaluating translation quality without references, but this restricts diagnosis to specific types of errors, such as incorrect translation of single numeric or currency words. In order to diagnose general errors, this paper proposes a new Bilingual Translation Pair Generation based Behavior Testing (BTPGBT) framework for conducting behavioral testing of MT systems. The core idea of BTPGBT is to employ a novel bilingual translation pair generation (BTPG) approach that automates the construction of high-quality test cases and their pseudoreferences. Experimental results on various MT systems demonstrate that BTPGBT could provide comprehensive and accurate behavioral testing results for general error diagnosis, which further leads to several insightful findings. Our code and data are available at https: //github.com/wujunjie1998/BTPGBT. | [
"Wu, Junjie",
"Liu, Lemao",
"Yeung, Dit-Yan"
] | Towards General Error Diagnosis via Behavioral Testing in Machine Translation | findings-emnlp.446 | 2310.13362 | [
"https://github.com/wujunjie1998/btpgbt"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.447.bib | https://aclanthology.org/2023.findings-emnlp.447/ | @inproceedings{yu-etal-2023-retrieval,
title = "Retrieval-Augmented Few-shot Text Classification",
author = "Yu, Guoxin and
Liu, Lemao and
Jiang, Haiyun and
Shi, Shuming and
Ao, Xiang",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.447",
doi = "10.18653/v1/2023.findings-emnlp.447",
pages = "6721--6735",
abstract = "Retrieval-augmented methods are successful in the standard scenario where the retrieval space is sufficient; whereas in the few-shot scenario with limited retrieval space, this paper shows it is non-trivial to put them into practice. First, it is impossible to retrieve semantically similar examples by using an off-the-shelf metric and it is crucial to learn a task-specific retrieval metric; Second, our preliminary experiments demonstrate that it is difficult to optimize a plausible metric by minimizing the standard cross-entropy loss. The in-depth analyses quantitatively show minimizing cross-entropy loss suffers from the weak supervision signals and the severe gradient vanishing issue during the optimization. To address these issues, we introduce two novel training objectives, namely EM-L and R-L, which provide more task-specific guidance to the retrieval metric by the EM algorithm and a ranking-based loss, respectively. Extensive experiments on 10 datasets prove the superiority of the proposed retrieval augmented methods on the performance.",
}
| Retrieval-augmented methods are successful in the standard scenario where the retrieval space is sufficient; whereas in the few-shot scenario with limited retrieval space, this paper shows it is non-trivial to put them into practice. First, it is impossible to retrieve semantically similar examples by using an off-the-shelf metric and it is crucial to learn a task-specific retrieval metric; Second, our preliminary experiments demonstrate that it is difficult to optimize a plausible metric by minimizing the standard cross-entropy loss. The in-depth analyses quantitatively show minimizing cross-entropy loss suffers from the weak supervision signals and the severe gradient vanishing issue during the optimization. To address these issues, we introduce two novel training objectives, namely EM-L and R-L, which provide more task-specific guidance to the retrieval metric by the EM algorithm and a ranking-based loss, respectively. Extensive experiments on 10 datasets prove the superiority of the proposed retrieval augmented methods on the performance. | [
"Yu, Guoxin",
"Liu, Lemao",
"Jiang, Haiyun",
"Shi, Shuming",
"Ao, Xiang"
] | Retrieval-Augmented Few-shot Text Classification | findings-emnlp.447 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.448.bib | https://aclanthology.org/2023.findings-emnlp.448/ | @inproceedings{chen-etal-2023-temporal,
title = "Temporal Extrapolation and Knowledge Transfer for Lifelong Temporal Knowledge Graph Reasoning",
author = "Chen, Zhongwu and
Xu, Chengjin and
Su, Fenglong and
Huang, Zhen and
Dou, Yong",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.448",
doi = "10.18653/v1/2023.findings-emnlp.448",
pages = "6736--6746",
abstract = "Real-world Temporal Knowledge Graphs keep growing with time and new entities and facts emerge continually, necessitating a model that can extrapolate to future timestamps and transfer knowledge for new components. Therefore, our work first dives into this more realistic issue, lifelong TKG reasoning, where existing methods can only address part of the challenges. Specifically, we formulate lifelong TKG reasoning as a temporal-path-based reinforcement learning (RL) framework. Then, we add temporal displacement into the action space of RL to extrapolate for the future and further propose a temporal-rule-based reward shaping to guide the training. To transfer and update knowledge, we design a new edge-aware message passing module, where the embeddings of new entities and edges are inductive. We conduct extensive experiments on three newly constructed benchmarks for lifelong TKG reasoning. Experimental results show the outperforming effectiveness of our model against all well-adapted baselines.",
}
| Real-world Temporal Knowledge Graphs keep growing with time and new entities and facts emerge continually, necessitating a model that can extrapolate to future timestamps and transfer knowledge for new components. Therefore, our work first dives into this more realistic issue, lifelong TKG reasoning, where existing methods can only address part of the challenges. Specifically, we formulate lifelong TKG reasoning as a temporal-path-based reinforcement learning (RL) framework. Then, we add temporal displacement into the action space of RL to extrapolate for the future and further propose a temporal-rule-based reward shaping to guide the training. To transfer and update knowledge, we design a new edge-aware message passing module, where the embeddings of new entities and edges are inductive. We conduct extensive experiments on three newly constructed benchmarks for lifelong TKG reasoning. Experimental results show the outperforming effectiveness of our model against all well-adapted baselines. | [
"Chen, Zhongwu",
"Xu, Chengjin",
"Su, Fenglong",
"Huang, Zhen",
"Dou, Yong"
] | Temporal Extrapolation and Knowledge Transfer for Lifelong Temporal Knowledge Graph Reasoning | findings-emnlp.448 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.449.bib | https://aclanthology.org/2023.findings-emnlp.449/ | @inproceedings{ullah-etal-2023-comparing,
title = "Comparing Prompt-Based and Standard Fine-Tuning for {U}rdu Text Classification",
author = "Ullah, Faizad and
Azam, Ubaid and
Faheem, Ali and
Kamiran, Faisal and
Karim, Asim",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.449",
doi = "10.18653/v1/2023.findings-emnlp.449",
pages = "6747--6754",
abstract = "Recent advancements in natural language processing have demonstrated the efficacy of pre-trained language models for various downstream tasks through prompt-based fine-tuning. In contrast to standard fine-tuning, which relies solely on labeled examples, prompt-based fine-tuning combines a few labeled examples (few shot) with guidance through prompts tailored for the specific language and task. For low-resource languages, where labeled examples are limited, prompt-based fine-tuning appears to be a promising alternative. In this paper, we compare prompt-based and standard fine-tuning for the popular task of text classification in Urdu and Roman Urdu languages. We conduct experiments using five datasets, covering different domains, and pre-trained multilingual transformers. The results reveal that significant improvement of up to 13{\%} in accuracy is achieved by prompt-based fine-tuning over standard fine-tuning approaches. This suggests the potential of prompt-based fine-tuning as a valuable approach for low-resource languages with limited labeled data.",
}
| Recent advancements in natural language processing have demonstrated the efficacy of pre-trained language models for various downstream tasks through prompt-based fine-tuning. In contrast to standard fine-tuning, which relies solely on labeled examples, prompt-based fine-tuning combines a few labeled examples (few shot) with guidance through prompts tailored for the specific language and task. For low-resource languages, where labeled examples are limited, prompt-based fine-tuning appears to be a promising alternative. In this paper, we compare prompt-based and standard fine-tuning for the popular task of text classification in Urdu and Roman Urdu languages. We conduct experiments using five datasets, covering different domains, and pre-trained multilingual transformers. The results reveal that significant improvement of up to 13{\%} in accuracy is achieved by prompt-based fine-tuning over standard fine-tuning approaches. This suggests the potential of prompt-based fine-tuning as a valuable approach for low-resource languages with limited labeled data. | [
"Ullah, Faizad",
"Azam, Ubaid",
"Faheem, Ali",
"Kamiran, Faisal",
"Karim, Asim"
] | Comparing Prompt-Based and Standard Fine-Tuning for Urdu Text Classification | findings-emnlp.449 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.450.bib | https://aclanthology.org/2023.findings-emnlp.450/ | @inproceedings{son-etal-2023-explore,
title = "Explore the Way: Exploring Reasoning Path by Bridging Entities for Effective Cross-Document Relation Extraction",
author = "Son, Junyoung and
Kim, Jinsung and
Lim, Jungwoo and
Jang, Yoonna and
Lim, Heuiseok",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.450",
doi = "10.18653/v1/2023.findings-emnlp.450",
pages = "6755--6761",
abstract = "Cross-document relation extraction (CodRED) task aims to infer the relation between two entities mentioned in different documents within a reasoning path. Previous studies have concentrated on merely capturing implicit relations between the entities. However, humans usually utilize explicit information chains such as hyperlinks or additional searches to find the relations between two entities. Inspired by this, we propose Path wIth expLOraTion (PILOT) that provides the enhanced reasoning path by exploring the explicit clue information within the documents. PILOT finds the bridging entities which directly guide the paths between the entities and then employs them as stepstones to navigate desirable paths. We show that models with PILOT outperform the baselines in the CodRED task. Furthermore, we offer a variety of analyses to verify the validity of the reasoning paths constructed through PILOT, including evaluations using large language models such as ChatGPT.",
}
| Cross-document relation extraction (CodRED) task aims to infer the relation between two entities mentioned in different documents within a reasoning path. Previous studies have concentrated on merely capturing implicit relations between the entities. However, humans usually utilize explicit information chains such as hyperlinks or additional searches to find the relations between two entities. Inspired by this, we propose Path wIth expLOraTion (PILOT) that provides the enhanced reasoning path by exploring the explicit clue information within the documents. PILOT finds the bridging entities which directly guide the paths between the entities and then employs them as stepstones to navigate desirable paths. We show that models with PILOT outperform the baselines in the CodRED task. Furthermore, we offer a variety of analyses to verify the validity of the reasoning paths constructed through PILOT, including evaluations using large language models such as ChatGPT. | [
"Son, Junyoung",
"Kim, Jinsung",
"Lim, Jungwoo",
"Jang, Yoonna",
"Lim, Heuiseok"
] | Explore the Way: Exploring Reasoning Path by Bridging Entities for Effective Cross-Document Relation Extraction | findings-emnlp.450 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.451.bib | https://aclanthology.org/2023.findings-emnlp.451/ | @inproceedings{ashok-etal-2023-student,
title = "The student becomes the master: Outperforming {GPT}3 on Scientific Factual Error Correction",
author = "Ashok, Dhananjay and
Kulkarni, Atharva and
Pham, Hai and
Poczos, Barnabas",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.451",
doi = "10.18653/v1/2023.findings-emnlp.451",
pages = "6762--6778",
abstract = "Due to the prohibitively high cost of creating error correction datasets, most Factual Claim Correction methods rely on a powerful verification model to guide the correction process. This leads to a significant drop in performance in domains like Scientific Claim Correction, where good verification models do not always exist. In this work we introduce SciFix, a claim correction system that does not require a verifier but is able to outperform existing methods by a considerable margin {---} achieving correction accuracy of 84{\%} on the SciFact dataset, 77{\%} on SciFact-Open and 72.75{\%} on the CovidFact dataset, compared to next best accuracies of 7.6{\%}, 5{\%} and 15{\%} on the same datasets respectively. Our method leverages the power of prompting with LLMs during training to create a richly annotated dataset that can be used for fully supervised training and regularization. We additionally use a claim-aware decoding procedure to improve the quality of corrected claims. Our method outperforms the very LLM that was used to generate the annotated dataset {---} with FewShot Prompting on GPT3.5 achieving 58{\%}, 61{\%} and 64{\%} on the respective datasets, a consistently lower correction accuracy, despite using nearly 800 times as many parameters as our model.",
}
| Due to the prohibitively high cost of creating error correction datasets, most Factual Claim Correction methods rely on a powerful verification model to guide the correction process. This leads to a significant drop in performance in domains like Scientific Claim Correction, where good verification models do not always exist. In this work we introduce SciFix, a claim correction system that does not require a verifier but is able to outperform existing methods by a considerable margin {---} achieving correction accuracy of 84{\%} on the SciFact dataset, 77{\%} on SciFact-Open and 72.75{\%} on the CovidFact dataset, compared to next best accuracies of 7.6{\%}, 5{\%} and 15{\%} on the same datasets respectively. Our method leverages the power of prompting with LLMs during training to create a richly annotated dataset that can be used for fully supervised training and regularization. We additionally use a claim-aware decoding procedure to improve the quality of corrected claims. Our method outperforms the very LLM that was used to generate the annotated dataset {---} with FewShot Prompting on GPT3.5 achieving 58{\%}, 61{\%} and 64{\%} on the respective datasets, a consistently lower correction accuracy, despite using nearly 800 times as many parameters as our model. | [
"Ashok, Dhananjay",
"Kulkarni, Atharva",
"Pham, Hai",
"Poczos, Barnabas"
] | The student becomes the master: Outperforming GPT3 on Scientific Factual Error Correction | findings-emnlp.451 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.452.bib | https://aclanthology.org/2023.findings-emnlp.452/ | @inproceedings{li-du-2023-leveraging,
title = "Leveraging Structured Information for Explainable Multi-hop Question Answering and Reasoning",
author = "Li, Ruosen and
Du, Xinya",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.452",
doi = "10.18653/v1/2023.findings-emnlp.452",
pages = "6779--6789",
abstract = "Neural models, including large language models (LLMs), achieve superior performance on multi-hop question-answering. To elicit reasoning capabilities from LLMs, recent works propose using the chain-of-thought (CoT) mechanism to generate both the reasoning chain and the answer, which enhances the model{'}s capabilities in conducting multi-hop reasoning. However, several challenges still remain: such as struggling with inaccurate reasoning, hallucinations, and lack of interpretability. On the other hand, information extraction (IE) identifies entities, relations, and events grounded to the text. The extracted structured information can be easily interpreted by humans and machines (Grishman, 2019). In this work, we investigate constructing and leveraging extracted semantic structures (graphs) for multi-hop question answering, especially the reasoning process. Empirical results and human evaluations show that our framework: generates more faithful reasoning chains and substantially improves the QA performance on two benchmark datasets. Moreover, the extracted structures themselves naturally provide grounded explanations that are preferred by humans, as compared to the generated reasoning chains and saliency-based explanations.",
}
| Neural models, including large language models (LLMs), achieve superior performance on multi-hop question-answering. To elicit reasoning capabilities from LLMs, recent works propose using the chain-of-thought (CoT) mechanism to generate both the reasoning chain and the answer, which enhances the model{'}s capabilities in conducting multi-hop reasoning. However, several challenges still remain: such as struggling with inaccurate reasoning, hallucinations, and lack of interpretability. On the other hand, information extraction (IE) identifies entities, relations, and events grounded to the text. The extracted structured information can be easily interpreted by humans and machines (Grishman, 2019). In this work, we investigate constructing and leveraging extracted semantic structures (graphs) for multi-hop question answering, especially the reasoning process. Empirical results and human evaluations show that our framework: generates more faithful reasoning chains and substantially improves the QA performance on two benchmark datasets. Moreover, the extracted structures themselves naturally provide grounded explanations that are preferred by humans, as compared to the generated reasoning chains and saliency-based explanations. | [
"Li, Ruosen",
"Du, Xinya"
] | Leveraging Structured Information for Explainable Multi-hop Question Answering and Reasoning | findings-emnlp.452 | 2311.03734 | [
"https://github.com/bcdnlp/structure-qa"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.453.bib | https://aclanthology.org/2023.findings-emnlp.453/ | @inproceedings{zhu-etal-2023-hierarchical,
title = "Hierarchical Catalogue Generation for Literature Review: A Benchmark",
author = "Zhu, Kun and
Feng, Xiaocheng and
Feng, Xiachong and
Wu, Yingsheng and
Qin, Bing",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.453",
doi = "10.18653/v1/2023.findings-emnlp.453",
pages = "6790--6804",
abstract = "Scientific literature review generation aims to extract and organize important information from an abundant collection of reference papers and produces corresponding reviews while lacking a clear and logical hierarchy. We observe that a high-quality catalogue-guided generation process can effectively alleviate this problem. Therefore, we present an atomic and challenging task named Hierarchical Catalogue Generation for Literature Review as the first step for review generation, which aims to produce a hierarchical catalogue of a review paper given various references. We construct a novel English Hierarchical Catalogues of Literature Reviews Dataset with 7.6k literature review catalogues and 389k reference papers. To accurately assess the model performance, we design two evaluation metrics for informativeness and similarity to ground truth from semantics and structure. Our extensive analyses verify the high quality of our dataset and the effectiveness of our evaluation metrics. We further benchmark diverse experiments on state-of-the-art summarization models like BART and large language models like ChatGPT to evaluate their capabilities. We further discuss potential directions for this task to motivate future research.",
}
| Scientific literature review generation aims to extract and organize important information from an abundant collection of reference papers and produces corresponding reviews while lacking a clear and logical hierarchy. We observe that a high-quality catalogue-guided generation process can effectively alleviate this problem. Therefore, we present an atomic and challenging task named Hierarchical Catalogue Generation for Literature Review as the first step for review generation, which aims to produce a hierarchical catalogue of a review paper given various references. We construct a novel English Hierarchical Catalogues of Literature Reviews Dataset with 7.6k literature review catalogues and 389k reference papers. To accurately assess the model performance, we design two evaluation metrics for informativeness and similarity to ground truth from semantics and structure. Our extensive analyses verify the high quality of our dataset and the effectiveness of our evaluation metrics. We further benchmark diverse experiments on state-of-the-art summarization models like BART and large language models like ChatGPT to evaluate their capabilities. We further discuss potential directions for this task to motivate future research. | [
"Zhu, Kun",
"Feng, Xiaocheng",
"Feng, Xiachong",
"Wu, Yingsheng",
"Qin, Bing"
] | Hierarchical Catalogue Generation for Literature Review: A Benchmark | findings-emnlp.453 | 2304.03512 | [
"https://github.com/zhukun1020/hicad"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.454.bib | https://aclanthology.org/2023.findings-emnlp.454/ | @inproceedings{chen-etal-2023-mcc,
title = "{MCC}-{KD}: Multi-{C}o{T} Consistent Knowledge Distillation",
author = "Chen, Hongzhan and
Wu, Siyue and
Quan, Xiaojun and
Wang, Rui and
Yan, Ming and
Zhang, Ji",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.454",
doi = "10.18653/v1/2023.findings-emnlp.454",
pages = "6805--6820",
abstract = "Large language models (LLMs) have showcased remarkable capabilities in complex reasoning through chain of thought (CoT) prompting. Recently, there has been a growing interest in transferring these reasoning abilities from LLMs to smaller models. However, achieving both the diversity and consistency in rationales presents a challenge. In this paper, we focus on enhancing these two aspects and propose Multi-CoT Consistent Knowledge Distillation (MCC-KD) to efficiently distill the reasoning capabilities. In MCC-KD, we generate multiple rationales for each question and enforce consistency among their predictions by minimizing the bidirectional KL-divergence between the answer distributions. We conduct comprehensive experiments to investigate the effectiveness of MCC-KD with different model architectures (LLaMA/FlanT5) and various model scales (3B/7B/11B/13B) on both mathematical reasoning and commonsense reasoning benchmarks. The empirical results demonstrate that MCC-KD achieves superior performance on in-distribution datasets and exhibits a strong generalization ability on out-of-distribution datasets.",
}
| Large language models (LLMs) have showcased remarkable capabilities in complex reasoning through chain of thought (CoT) prompting. Recently, there has been a growing interest in transferring these reasoning abilities from LLMs to smaller models. However, achieving both the diversity and consistency in rationales presents a challenge. In this paper, we focus on enhancing these two aspects and propose Multi-CoT Consistent Knowledge Distillation (MCC-KD) to efficiently distill the reasoning capabilities. In MCC-KD, we generate multiple rationales for each question and enforce consistency among their predictions by minimizing the bidirectional KL-divergence between the answer distributions. We conduct comprehensive experiments to investigate the effectiveness of MCC-KD with different model architectures (LLaMA/FlanT5) and various model scales (3B/7B/11B/13B) on both mathematical reasoning and commonsense reasoning benchmarks. The empirical results demonstrate that MCC-KD achieves superior performance on in-distribution datasets and exhibits a strong generalization ability on out-of-distribution datasets. | [
"Chen, Hongzhan",
"Wu, Siyue",
"Quan, Xiaojun",
"Wang, Rui",
"Yan, Ming",
"Zhang, Ji"
] | MCC-KD: Multi-CoT Consistent Knowledge Distillation | findings-emnlp.454 | 2310.14747 | [
"https://github.com/homzer/MCC-KD"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.455.bib | https://aclanthology.org/2023.findings-emnlp.455/ | @inproceedings{wu-etal-2023-empirical,
title = "An Empirical Study of Frame Selection for Text-to-Video Retrieval",
author = "Wu, Mengxia and
Cao, Min and
Bai, Yang and
Zeng, Ziyin and
Chen, Chen and
Nie, Liqiang and
Zhang, Min",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.455",
doi = "10.18653/v1/2023.findings-emnlp.455",
pages = "6821--6832",
abstract = "Text-to-video retrieval (TVR) aims to find the most relevant video in a large video gallery given a query text. The intricate and abundant context of the video challenges the performance and efficiency of TVR. To handle the serialized video contexts, existing methods typically select a subset of frames within a video to represent the video content for TVR. How to select the most representative frames is a crucial issue, whereby the selected frames are required to not only retain the semantic information of the video but also promote retrieval efficiency by excluding temporally redundant frames. In this paper, we make the first empirical study of frame selection for TVR. We systemically classify existing frame selection methods into text-free and text-guided ones, under which we detailedly analyze six different frame selections in terms of effectiveness and efficiency. Among them, two frame selections are first developed in this paper. According to the comprehensive analysis on multiple TVR benchmarks, we empirically conclude that the TVR with proper frame selections can significantly improve the retrieval efficiency without sacrificing the retrieval performance.",
}
| Text-to-video retrieval (TVR) aims to find the most relevant video in a large video gallery given a query text. The intricate and abundant context of the video challenges the performance and efficiency of TVR. To handle the serialized video contexts, existing methods typically select a subset of frames within a video to represent the video content for TVR. How to select the most representative frames is a crucial issue, whereby the selected frames are required to not only retain the semantic information of the video but also promote retrieval efficiency by excluding temporally redundant frames. In this paper, we make the first empirical study of frame selection for TVR. We systemically classify existing frame selection methods into text-free and text-guided ones, under which we detailedly analyze six different frame selections in terms of effectiveness and efficiency. Among them, two frame selections are first developed in this paper. According to the comprehensive analysis on multiple TVR benchmarks, we empirically conclude that the TVR with proper frame selections can significantly improve the retrieval efficiency without sacrificing the retrieval performance. | [
"Wu, Mengxia",
"Cao, Min",
"Bai, Yang",
"Zeng, Ziyin",
"Chen, Chen",
"Nie, Liqiang",
"Zhang, Min"
] | An Empirical Study of Frame Selection for Text-to-Video Retrieval | findings-emnlp.455 | 2311.00298 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.456.bib | https://aclanthology.org/2023.findings-emnlp.456/ | @inproceedings{kim-etal-2023-conditional,
title = "Conditional Natural Language Inference",
author = "Kim, Youngwoo and
Rahimi, Razieh and
Allan, James",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.456",
doi = "10.18653/v1/2023.findings-emnlp.456",
pages = "6833--6851",
abstract = "To properly explain sentence pairs that provide contradictory (different) information for different conditions, we introduce the task of conditional natural language inference (Cond-NLI) and focus on automatically extracting contradictory aspects and their conditions from a sentence pair. Cond-NLI can help to provide a full spectrum of information, such as when there are multiple answers to a question each addressing a specific condition, or reviews with different opinions for different conditions. We show that widely-used feature-attribution explanation models are not suitable for finding conditions, especially when sentences are long and are written independently. We propose a simple yet effective model for the original NLI task that can successfully extract conditions while not requiring token-level annotations. Our model enhances the interpretability of the NLI task while maintaining comparable accuracy. To evaluate models for the Cond-NLI, we build and release a token-level annotated dataset BioClaim which contains potentially contradictory claims from the biomedical domain. Our experiments show that our proposed model outperforms the full cross-encoder and other baselines in extracting conditions. It also performs on-par with GPT-3 which has an order of magnitude more parameters and trained on a huge amount of data.",
}
| To properly explain sentence pairs that provide contradictory (different) information for different conditions, we introduce the task of conditional natural language inference (Cond-NLI) and focus on automatically extracting contradictory aspects and their conditions from a sentence pair. Cond-NLI can help to provide a full spectrum of information, such as when there are multiple answers to a question each addressing a specific condition, or reviews with different opinions for different conditions. We show that widely-used feature-attribution explanation models are not suitable for finding conditions, especially when sentences are long and are written independently. We propose a simple yet effective model for the original NLI task that can successfully extract conditions while not requiring token-level annotations. Our model enhances the interpretability of the NLI task while maintaining comparable accuracy. To evaluate models for the Cond-NLI, we build and release a token-level annotated dataset BioClaim which contains potentially contradictory claims from the biomedical domain. Our experiments show that our proposed model outperforms the full cross-encoder and other baselines in extracting conditions. It also performs on-par with GPT-3 which has an order of magnitude more parameters and trained on a huge amount of data. | [
"Kim, Youngwoo",
"Rahimi, Razieh",
"Allan, James"
] | Conditional Natural Language Inference | findings-emnlp.456 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.457.bib | https://aclanthology.org/2023.findings-emnlp.457/ | @inproceedings{bian-etal-2023-contrastive,
title = "Contrastive Distant Supervision for Debiased and Denoised Machine Reading Comprehension",
author = "Bian, Ning and
Lin, Hongyu and
Han, Xianpei and
He, Ben and
Sun, Le",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.457",
doi = "10.18653/v1/2023.findings-emnlp.457",
pages = "6852--6863",
abstract = "Distant Supervision (DS) is a promising learning approach for MRC by leveraging easily-obtained question-answer pairs. Unfortunately, the heuristically annotated dataset will inevitably lead to mislabeled instances, resulting in answer bias and context noise problems. To learn debiased and denoised MRC models, this paper proposes the Contrastive Distant Supervision algorithm {--} CDS, which can learn to distinguish confusing and noisy instances via confidence-aware contrastive learning. Specifically, to eliminate answer bias, CDS samples counterfactual negative instances, which ensures that MRC models must take both answer information and question-context interaction into consideration. To denoise distantly annotated contexts, CDS samples confusing negative instances to increase the margin between correct and mislabeled instances. We further propose a confidence-aware contrastive loss to model and leverage the uncertainty of all DS instances during learning. Experimental results show that CDS is effective and can even outperform supervised MRC models without manual annotations.",
}
| Distant Supervision (DS) is a promising learning approach for MRC by leveraging easily-obtained question-answer pairs. Unfortunately, the heuristically annotated dataset will inevitably lead to mislabeled instances, resulting in answer bias and context noise problems. To learn debiased and denoised MRC models, this paper proposes the Contrastive Distant Supervision algorithm {--} CDS, which can learn to distinguish confusing and noisy instances via confidence-aware contrastive learning. Specifically, to eliminate answer bias, CDS samples counterfactual negative instances, which ensures that MRC models must take both answer information and question-context interaction into consideration. To denoise distantly annotated contexts, CDS samples confusing negative instances to increase the margin between correct and mislabeled instances. We further propose a confidence-aware contrastive loss to model and leverage the uncertainty of all DS instances during learning. Experimental results show that CDS is effective and can even outperform supervised MRC models without manual annotations. | [
"Bian, Ning",
"Lin, Hongyu",
"Han, Xianpei",
"He, Ben",
"Sun, Le"
] | Contrastive Distant Supervision for Debiased and Denoised Machine Reading Comprehension | findings-emnlp.457 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.458.bib | https://aclanthology.org/2023.findings-emnlp.458/ | @inproceedings{li-etal-2023-keplet,
title = "{KEPLET}: Knowledge-Enhanced Pretrained Language Model with Topic Entity Awareness",
author = "Li, Yichuan and
Han, Jialong and
Lee, Kyumin and
Ma, Chengyuan and
Yao, Benjamin and
Liu, Xiaohu",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.458",
doi = "10.18653/v1/2023.findings-emnlp.458",
pages = "6864--6876",
abstract = "In recent years, Pre-trained Language Models (PLMs) have shown their superiority by pre-training on unstructured text corpus and then fine-tuning on downstream tasks. On entity-rich textual resources like Wikipedia, Knowledge-Enhanced PLMs (KEPLMs) incorporate the interactions between tokens and mentioned entities in pre-training, and are thus more effective on entity-centric tasks such as entity linking and relation classification. Although exploiting Wikipedia{'}s rich structures to some extent, conventional KEPLMs still neglect a unique layout of the corpus where each Wikipedia page is around a topic entity (identified by the page URL and shown in the page title). In this paper, we demonstrate that KEPLMs without incorporating the topic entities will lead to insufficient entity interaction and biased (relation) word semantics. We thus propose KEPLET, a novel Knowledge-{\'E}nhanced Pre-trained LanguagE model with Topic entity awareness. In an end-to-end manner, KEPLET identifies where to add the topic entity{'}s information in a Wikipedia sentence, fuses such information into token and mentioned entities representations, and supervises the network learning, through which it takes topic entities back into consideration. Experiments demonstrated the generality and superiority of KEPLET which was applied to two representative KEPLMs, achieving significant improvements on four entity-centric tasks.",
}
| In recent years, Pre-trained Language Models (PLMs) have shown their superiority by pre-training on unstructured text corpus and then fine-tuning on downstream tasks. On entity-rich textual resources like Wikipedia, Knowledge-Enhanced PLMs (KEPLMs) incorporate the interactions between tokens and mentioned entities in pre-training, and are thus more effective on entity-centric tasks such as entity linking and relation classification. Although exploiting Wikipedia{'}s rich structures to some extent, conventional KEPLMs still neglect a unique layout of the corpus where each Wikipedia page is around a topic entity (identified by the page URL and shown in the page title). In this paper, we demonstrate that KEPLMs without incorporating the topic entities will lead to insufficient entity interaction and biased (relation) word semantics. We thus propose KEPLET, a novel Knowledge-{\'E}nhanced Pre-trained LanguagE model with Topic entity awareness. In an end-to-end manner, KEPLET identifies where to add the topic entity{'}s information in a Wikipedia sentence, fuses such information into token and mentioned entities representations, and supervises the network learning, through which it takes topic entities back into consideration. Experiments demonstrated the generality and superiority of KEPLET which was applied to two representative KEPLMs, achieving significant improvements on four entity-centric tasks. | [
"Li, Yichuan",
"Han, Jialong",
"Lee, Kyumin",
"Ma, Chengyuan",
"Yao, Benjamin",
"Liu, Xiaohu"
] | KEPLET: Knowledge-Enhanced Pretrained Language Model with Topic Entity Awareness | findings-emnlp.458 | 2305.01810 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.459.bib | https://aclanthology.org/2023.findings-emnlp.459/ | @inproceedings{li-etal-2023-revisiting-large,
title = "Revisiting Large Language Models as Zero-shot Relation Extractors",
author = "Li, Guozheng and
Wang, Peng and
Ke, Wenjun",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.459",
pages = "6877--6892",
abstract = "Relation extraction (RE) consistently involves a certain degree of labeled or unlabeled data even if under zero-shot setting. Recent studies have shown that large language models (LLMs) transfer well to new tasks out-of-the-box simply given a natural language prompt, which provides the possibility of extracting relations from text without any data and parameter tuning. This work focuses on the study of exploring LLMs, such as ChatGPT, as zero-shot relation extractors. On the one hand, we analyze the drawbacks of existing RE prompts and attempt to incorporate recent prompt techniques such as chain-of-thought (CoT) to improve zero-shot RE. We propose the summarize-and-ask (SumAsk) prompting, a simple prompt recursively using LLMs to transform RE inputs to the effective question answering (QA) format. On the other hand, we conduct comprehensive experiments on various benchmarks and settings to investigate the capabilities of LLMs on zero-shot RE. Specifically, we have the following findings: (i) SumAsk consistently and significantly improves LLMs performance on different model sizes, benchmarks and settings; (ii) Zero-shot prompting with ChatGPT achieves competitive or superior results compared with zero-shot and fully supervised methods; (iii) LLMs deliver promising performance in extracting overlapping relations; (iv) The performance varies greatly regarding different relations. Different from small language models, LLMs are effective in handling challenge none-of-the-above (NoTA) relation.",
}
| Relation extraction (RE) consistently involves a certain degree of labeled or unlabeled data even if under zero-shot setting. Recent studies have shown that large language models (LLMs) transfer well to new tasks out-of-the-box simply given a natural language prompt, which provides the possibility of extracting relations from text without any data and parameter tuning. This work focuses on the study of exploring LLMs, such as ChatGPT, as zero-shot relation extractors. On the one hand, we analyze the drawbacks of existing RE prompts and attempt to incorporate recent prompt techniques such as chain-of-thought (CoT) to improve zero-shot RE. We propose the summarize-and-ask (SumAsk) prompting, a simple prompt recursively using LLMs to transform RE inputs to the effective question answering (QA) format. On the other hand, we conduct comprehensive experiments on various benchmarks and settings to investigate the capabilities of LLMs on zero-shot RE. Specifically, we have the following findings: (i) SumAsk consistently and significantly improves LLMs performance on different model sizes, benchmarks and settings; (ii) Zero-shot prompting with ChatGPT achieves competitive or superior results compared with zero-shot and fully supervised methods; (iii) LLMs deliver promising performance in extracting overlapping relations; (iv) The performance varies greatly regarding different relations. Different from small language models, LLMs are effective in handling challenge none-of-the-above (NoTA) relation. | [
"Li, Guozheng",
"Wang, Peng",
"Ke, Wenjun"
] | Revisiting Large Language Models as Zero-shot Relation Extractors | findings-emnlp.459 | 2310.05028 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.460.bib | https://aclanthology.org/2023.findings-emnlp.460/ | @inproceedings{zhou-etal-2023-multi,
title = "Multi-Stage Pre-training Enhanced by {C}hat{GPT} for Multi-Scenario Multi-Domain Dialogue Summarization",
author = "Zhou, Weixiao and
Li, Gengyao and
Cheng, Xianfu and
Liang, Xinnian and
Zhu, Junnan and
Zhai, Feifei and
Li, Zhoujun",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.460",
doi = "10.18653/v1/2023.findings-emnlp.460",
pages = "6893--6908",
abstract = "Dialogue summarization involves a wide range of scenarios and domains. However, existing methods generally only apply to specific scenarios or domains. In this study, we propose a new pre-trained model specifically designed for multi-scenario multi-domain dialogue summarization. It adopts a multi-stage pre-training strategy to reduce the gap between the pre-training objective and fine-tuning objective. Specifically, we first conduct domain-aware pre-training using large-scale multi-scenario multi-domain dialogue data to enhance the adaptability of our pre-trained model. Then, we conduct task-oriented pre-training using large-scale multi-scenario multi-domain {``}dialogue-summary{''} parallel data annotated by ChatGPT to enhance the dialogue summarization ability of our pre-trained model. Experimental results on three dialogue summarization datasets from different scenarios and domains indicate that our pre-trained model significantly outperforms previous state-of-the-art models in full fine-tuning, zero-shot, and few-shot settings.",
}
| Dialogue summarization involves a wide range of scenarios and domains. However, existing methods generally only apply to specific scenarios or domains. In this study, we propose a new pre-trained model specifically designed for multi-scenario multi-domain dialogue summarization. It adopts a multi-stage pre-training strategy to reduce the gap between the pre-training objective and fine-tuning objective. Specifically, we first conduct domain-aware pre-training using large-scale multi-scenario multi-domain dialogue data to enhance the adaptability of our pre-trained model. Then, we conduct task-oriented pre-training using large-scale multi-scenario multi-domain {``}dialogue-summary{''} parallel data annotated by ChatGPT to enhance the dialogue summarization ability of our pre-trained model. Experimental results on three dialogue summarization datasets from different scenarios and domains indicate that our pre-trained model significantly outperforms previous state-of-the-art models in full fine-tuning, zero-shot, and few-shot settings. | [
"Zhou, Weixiao",
"Li, Gengyao",
"Cheng, Xianfu",
"Liang, Xinnian",
"Zhu, Junnan",
"Zhai, Feifei",
"Li, Zhoujun"
] | Multi-Stage Pre-training Enhanced by ChatGPT for Multi-Scenario Multi-Domain Dialogue Summarization | findings-emnlp.460 | 2310.10285 | [
"https://github.com/zhouweixiao/mp4"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.461.bib | https://aclanthology.org/2023.findings-emnlp.461/ | @inproceedings{muthusamy-etal-2023-towards,
title = "Towards large language model-based personal agents in the enterprise: Current trends and open problems",
author = "Muthusamy, Vinod and
Rizk, Yara and
Kate, Kiran and
Venkateswaran, Praveen and
Isahagian, Vatche and
Gulati, Ashu and
Dube, Parijat",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.461",
doi = "10.18653/v1/2023.findings-emnlp.461",
pages = "6909--6921",
abstract = "There is an emerging trend to use large language models (LLMs) to reason about complex goals and orchestrate a set of pluggable tools or APIs to accomplish a goal. This functionality could, among other use cases, be used to build personal assistants for knowledge workers. While there are impressive demos of LLMs being used as autonomous agents or for tool composition, these solutions are not ready mission-critical enterprise settings. For example, they are brittle to input changes, and can produce inconsistent results for the same inputs. These use cases have many open problems in an exciting area of NLP research, such as trust and explainability, consistency and reproducibility, adherence to guardrails and policies, best practices for composable tool design, and the need for new metrics and benchmarks. This vision paper illustrates some examples of LLM-based autonomous agents that reason and compose tools, highlights cases where they fail, surveys some of the recent efforts in this space, and lays out the research challenges to make these solutions viable for enterprises.",
}
| There is an emerging trend to use large language models (LLMs) to reason about complex goals and orchestrate a set of pluggable tools or APIs to accomplish a goal. This functionality could, among other use cases, be used to build personal assistants for knowledge workers. While there are impressive demos of LLMs being used as autonomous agents or for tool composition, these solutions are not ready mission-critical enterprise settings. For example, they are brittle to input changes, and can produce inconsistent results for the same inputs. These use cases have many open problems in an exciting area of NLP research, such as trust and explainability, consistency and reproducibility, adherence to guardrails and policies, best practices for composable tool design, and the need for new metrics and benchmarks. This vision paper illustrates some examples of LLM-based autonomous agents that reason and compose tools, highlights cases where they fail, surveys some of the recent efforts in this space, and lays out the research challenges to make these solutions viable for enterprises. | [
"Muthusamy, Vinod",
"Rizk, Yara",
"Kate, Kiran",
"Venkateswaran, Praveen",
"Isahagian, Vatche",
"Gulati, Ashu",
"Dube, Parijat"
] | Towards large language model-based personal agents in the enterprise: Current trends and open problems | findings-emnlp.461 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.462.bib | https://aclanthology.org/2023.findings-emnlp.462/ | @inproceedings{qian-etal-2023-creator,
title = "{CREATOR}: Tool Creation for Disentangling Abstract and Concrete Reasoning of Large Language Models",
author = "Qian, Cheng and
Han, Chi and
Fung, Yi and
Qin, Yujia and
Liu, Zhiyuan and
Ji, Heng",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.462",
doi = "10.18653/v1/2023.findings-emnlp.462",
pages = "6922--6939",
abstract = "Large Language Models (LLMs) have made significant progress in utilizing tools, but their ability is limited by API availability and the instability of implicit reasoning, particularly when both planning and execution are involved. To overcome these limitations, we propose CREATOR, a novel framework that enables LLMs to create their own tools using documentation and code realization. CREATOR disentangles abstract tool creation and concrete decision execution, resulting in improved performance. We evaluate CREATOR on MATH and TabMWP benchmarks, respectively consisting of challenging math competition problems and diverse tabular contents. Remarkably, CREATOR outperforms existing chain-of-thought, program-of-thought, and tool-using baselines. Additionally, we introduce the Creation Challenge dataset, featuring 2K diverse questions, to emphasize the necessity and benefits of LLMs{'} tool creation ability. Further research demonstrates that leveraging LLMs as tool creators facilitates knowledge transfer, and LLMs exhibit varying levels of tool creation abilities, enabling them to adapt to diverse situations. The tool creation ability revolutionizes the LLM{'}s problem-solving paradigm, driving us closer to the next frontier of artificial intelligence.",
}
| Large Language Models (LLMs) have made significant progress in utilizing tools, but their ability is limited by API availability and the instability of implicit reasoning, particularly when both planning and execution are involved. To overcome these limitations, we propose CREATOR, a novel framework that enables LLMs to create their own tools using documentation and code realization. CREATOR disentangles abstract tool creation and concrete decision execution, resulting in improved performance. We evaluate CREATOR on MATH and TabMWP benchmarks, respectively consisting of challenging math competition problems and diverse tabular contents. Remarkably, CREATOR outperforms existing chain-of-thought, program-of-thought, and tool-using baselines. Additionally, we introduce the Creation Challenge dataset, featuring 2K diverse questions, to emphasize the necessity and benefits of LLMs{'} tool creation ability. Further research demonstrates that leveraging LLMs as tool creators facilitates knowledge transfer, and LLMs exhibit varying levels of tool creation abilities, enabling them to adapt to diverse situations. The tool creation ability revolutionizes the LLM{'}s problem-solving paradigm, driving us closer to the next frontier of artificial intelligence. | [
"Qian, Cheng",
"Han, Chi",
"Fung, Yi",
"Qin, Yujia",
"Liu, Zhiyuan",
"Ji, Heng"
] | CREATOR: Tool Creation for Disentangling Abstract and Concrete Reasoning of Large Language Models | findings-emnlp.462 | 2305.14318 | [
"https://github.com/qiancheng0/creator"
] | https://huggingface.co/papers/2305.14318 | 0 | 0 | 0 | 6 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.463.bib | https://aclanthology.org/2023.findings-emnlp.463/ | @inproceedings{maeda-etal-2023-query,
title = "Query-based Image Captioning from Multi-context 360$cdegree$ Images",
author = "Maeda, Koki and
Kurita, Shuhei and
Miyanishi, Taiki and
Okazaki, Naoaki",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.463",
doi = "10.18653/v1/2023.findings-emnlp.463",
pages = "6940--6954",
abstract = "A 360-degree image captures the entire scene without the limitations of a camera{'}s field of view, which makes it difficult to describe all the contexts in a single caption. We propose a novel task called Query-based Image Captioning (QuIC) for 360-degree images, where a query (words or short phrases) specifies the context to describe. This task is more challenging than the conventional image captioning task, which describes salient objects in images, as it requires fine-grained scene understanding to select the contents consistent with user{'}s intent based on the query. We construct a dataset for the new task that comprises 3,940 360-degree images and 18,459 pairs of queries and captions annotated manually. Experiments demonstrate that fine-tuning image captioning models further on our dataset can generate more diverse and controllable captions from multiple contexts of 360-degree images.",
}
| A 360-degree image captures the entire scene without the limitations of a camera{'}s field of view, which makes it difficult to describe all the contexts in a single caption. We propose a novel task called Query-based Image Captioning (QuIC) for 360-degree images, where a query (words or short phrases) specifies the context to describe. This task is more challenging than the conventional image captioning task, which describes salient objects in images, as it requires fine-grained scene understanding to select the contents consistent with user{'}s intent based on the query. We construct a dataset for the new task that comprises 3,940 360-degree images and 18,459 pairs of queries and captions annotated manually. Experiments demonstrate that fine-tuning image captioning models further on our dataset can generate more diverse and controllable captions from multiple contexts of 360-degree images. | [
"Maeda, Koki",
"Kurita, Shuhei",
"Miyanishi, Taiki",
"Okazaki, Naoaki"
] | Query-based Image Captioning from Multi-context 360cdegree Images | findings-emnlp.463 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.464.bib | https://aclanthology.org/2023.findings-emnlp.464/ | @inproceedings{yang-etal-2023-auto,
title = "Auto Search Indexer for End-to-End Document Retrieval",
author = "Yang, Tianchi and
Song, Minghui and
Zhang, Zihan and
Huang, Haizhen and
Deng, Weiwei and
Sun, Feng and
Zhang, Qi",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.464",
doi = "10.18653/v1/2023.findings-emnlp.464",
pages = "6955--6970",
abstract = "Generative retrieval, which is a new advanced paradigm for document retrieval, has recently attracted research interests, since it encodes all documents into the model and directly generates the retrieved documents. However, its power is still underutilized since it heavily relies on the {``}preprocessed{''} document identifiers (docids), thus limiting its retrieval performance and ability to retrieve new documents. In this paper, we propose a novel fully end-to-end retrieval paradigm. It can not only end-to-end learn the best docids for existing and new documents automatically via a semantic indexing module, but also perform end-to-end document retrieval via an encoder-decoder-based generative model, namely Auto Search Indexer (ASI). Besides, we design a reparameterization mechanism to combine the above two modules into a joint optimization framework. Extensive experimental results demonstrate the superiority of our model over advanced baselines on both public and industrial datasets and also verify the ability to deal with new documents.",
}
| Generative retrieval, which is a new advanced paradigm for document retrieval, has recently attracted research interests, since it encodes all documents into the model and directly generates the retrieved documents. However, its power is still underutilized since it heavily relies on the {``}preprocessed{''} document identifiers (docids), thus limiting its retrieval performance and ability to retrieve new documents. In this paper, we propose a novel fully end-to-end retrieval paradigm. It can not only end-to-end learn the best docids for existing and new documents automatically via a semantic indexing module, but also perform end-to-end document retrieval via an encoder-decoder-based generative model, namely Auto Search Indexer (ASI). Besides, we design a reparameterization mechanism to combine the above two modules into a joint optimization framework. Extensive experimental results demonstrate the superiority of our model over advanced baselines on both public and industrial datasets and also verify the ability to deal with new documents. | [
"Yang, Tianchi",
"Song, Minghui",
"Zhang, Zihan",
"Huang, Haizhen",
"Deng, Weiwei",
"Sun, Feng",
"Zhang, Qi"
] | Auto Search Indexer for End-to-End Document Retrieval | findings-emnlp.464 | 2310.12455 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.465.bib | https://aclanthology.org/2023.findings-emnlp.465/ | @inproceedings{ghosh-caliskan-2023-person,
title = "{`}Person{'} == Light-skinned, Western Man, and Sexualization of Women of Color: Stereotypes in Stable Diffusion",
author = "Ghosh, Sourojit and
Caliskan, Aylin",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.465",
doi = "10.18653/v1/2023.findings-emnlp.465",
pages = "6971--6985",
abstract = "We study stereotypes embedded within one of the most popular text-to-image generators: Stable Diffusion. We answer the question: what stereotypes of gender and nationality/continental identity does Stable Diffusion display in the absence of such information i.e. what gender and nationality/continental identity is assigned to {`}a person,{'} or to {`}a person from Asia.{'} Using CLIP-cosine similarity for zero-shot classification of images generated by CLIP-based Stable Diffusion v2.1 verified by manual examination, we chronicle results from 136 prompts (50 results/prompt) of front-facing images of faces from 6 different continents, 27 countries and 3 genders. We observe how Stable Diffusion results of {`}a person{'} without any additional gender/nationality information correspond closest to images of men (avg. similarity 0.64) and least with persons of nonbinary gender (avg. similarity 0.41), and to persons from Europe/North America (avg. similarities 0.71 and 0.68, respectively) over Africa/Asia (avg. similarities 0.43 and 0.41, respectively), pointing towards Stable Diffusion having a concerning representation of personhood to be a European/North American man. We also show continental stereotypes and resultant harms e.g. a person from Oceania is deemed to be Australian/New Zealander (avg. similarities 0.77 and 0.74, respectively) over Papua New Guinean (avg. similarity 0.31), pointing to the erasure of Indigenous Oceanic peoples, who form a majority over descendants of colonizers both in Papua New Guinea and in Oceania overall. Finally, we unexpectedly observe a pattern of sexualization of women, specifically Latin American, Mexican, Indian and Egyptian women, confirmed through an NSFW detector and verified by manual examination. This demonstrates how Stable Diffusion perpetuates Western fetishization of women of color through objectification in media, which if left unchecked will worsen this stereotypical representation. All code and relevant data will be made publicly available.",
}
| We study stereotypes embedded within one of the most popular text-to-image generators: Stable Diffusion. We answer the question: what stereotypes of gender and nationality/continental identity does Stable Diffusion display in the absence of such information i.e. what gender and nationality/continental identity is assigned to {`}a person,{'} or to {`}a person from Asia.{'} Using CLIP-cosine similarity for zero-shot classification of images generated by CLIP-based Stable Diffusion v2.1 verified by manual examination, we chronicle results from 136 prompts (50 results/prompt) of front-facing images of faces from 6 different continents, 27 countries and 3 genders. We observe how Stable Diffusion results of {`}a person{'} without any additional gender/nationality information correspond closest to images of men (avg. similarity 0.64) and least with persons of nonbinary gender (avg. similarity 0.41), and to persons from Europe/North America (avg. similarities 0.71 and 0.68, respectively) over Africa/Asia (avg. similarities 0.43 and 0.41, respectively), pointing towards Stable Diffusion having a concerning representation of personhood to be a European/North American man. We also show continental stereotypes and resultant harms e.g. a person from Oceania is deemed to be Australian/New Zealander (avg. similarities 0.77 and 0.74, respectively) over Papua New Guinean (avg. similarity 0.31), pointing to the erasure of Indigenous Oceanic peoples, who form a majority over descendants of colonizers both in Papua New Guinea and in Oceania overall. Finally, we unexpectedly observe a pattern of sexualization of women, specifically Latin American, Mexican, Indian and Egyptian women, confirmed through an NSFW detector and verified by manual examination. This demonstrates how Stable Diffusion perpetuates Western fetishization of women of color through objectification in media, which if left unchecked will worsen this stereotypical representation. All code and relevant data will be made publicly available. | [
"Ghosh, Sourojit",
"Caliskan, Aylin"
] | `Person' == Light-skinned, Western Man, and Sexualization of Women of Color: Stereotypes in Stable Diffusion | findings-emnlp.465 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.466.bib | https://aclanthology.org/2023.findings-emnlp.466/ | @inproceedings{cai-etal-2023-task,
title = "Task-Attentive Transformer Architecture for Continual Learning of Vision-and-Language Tasks Using Knowledge Distillation",
author = "Cai, Yuliang and
Thomason, Jesse and
Rostami, Mohammad",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.466",
doi = "10.18653/v1/2023.findings-emnlp.466",
pages = "6986--7000",
abstract = "The size and the computational load of fine-tuning large-scale pre-trained neural network are becoming two major obstacles in adopting machine learning in many applications. Continual learning (CL) can serve as a remedy through enabling knowledge-transfer across sequentially arriving tasks which relaxes the need to fine-tune all network weights from scratch. However, existing CL algorithms primarily consider learning unimodal vision-only or language-only tasks. We develop a transformer-based CL architecture for learning bimodal vision-and-language tasks based on increasing the number of the learnable parameters dynamically and using knowledge distillation. The new additional parameters are used to specialize the network for each task. Our approach enables sharing information between the tasks while addressing the challenge of catastrophic forgetting. Our approach is scalable learning to a large number of tasks because it requires little memory and time overhead. Our model reaches state-of-the-art performance on challenging vision-and-language tasks.",
}
| The size and the computational load of fine-tuning large-scale pre-trained neural network are becoming two major obstacles in adopting machine learning in many applications. Continual learning (CL) can serve as a remedy through enabling knowledge-transfer across sequentially arriving tasks which relaxes the need to fine-tune all network weights from scratch. However, existing CL algorithms primarily consider learning unimodal vision-only or language-only tasks. We develop a transformer-based CL architecture for learning bimodal vision-and-language tasks based on increasing the number of the learnable parameters dynamically and using knowledge distillation. The new additional parameters are used to specialize the network for each task. Our approach enables sharing information between the tasks while addressing the challenge of catastrophic forgetting. Our approach is scalable learning to a large number of tasks because it requires little memory and time overhead. Our model reaches state-of-the-art performance on challenging vision-and-language tasks. | [
"Cai, Yuliang",
"Thomason, Jesse",
"Rostami, Mohammad"
] | Task-Attentive Transformer Architecture for Continual Learning of Vision-and-Language Tasks Using Knowledge Distillation | findings-emnlp.466 | 2303.14423 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.467.bib | https://aclanthology.org/2023.findings-emnlp.467/ | @inproceedings{liu-etal-2023-evaluating,
title = "Evaluating Verifiability in Generative Search Engines",
author = "Liu, Nelson and
Zhang, Tianyi and
Liang, Percy",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.467",
doi = "10.18653/v1/2023.findings-emnlp.467",
pages = "7001--7025",
abstract = "Generative search engines directly generate responses to user queries, along with in-line citations. A prerequisite trait of a trustworthy generative search engine is verifiability, i.e., systems should cite comprehensively (high citation recall; all statements are fully supported by citations) and accurately (high citation precision; every cite supports its associated statement). We conduct human evaluation to audit four popular generative search engines{---}Bing Chat, NeevaAI, perplexity.ai, and YouChat{---}across a diverse set of queries from a variety of sources (e.g., historical Google user queries, dynamically-collected open-ended questions on Reddit, etc.). We find that responses from existing generative search engines are fluent and appear informative, but frequently contain unsupported statements and inaccurate citations: on average, a mere 51.5{\%} of generated sentences are fully supported by citations and only 74.5{\%} of citations support their associated sentence. We believe that these results are concerningly low for systems that may serve as a primary tool for information-seeking users, especially given their facade of trustworthiness. We hope that our results further motivate the development of trustworthy generative search engines and help researchers and users better understand the shortcomings of existing commercial systems.",
}
| Generative search engines directly generate responses to user queries, along with in-line citations. A prerequisite trait of a trustworthy generative search engine is verifiability, i.e., systems should cite comprehensively (high citation recall; all statements are fully supported by citations) and accurately (high citation precision; every cite supports its associated statement). We conduct human evaluation to audit four popular generative search engines{---}Bing Chat, NeevaAI, perplexity.ai, and YouChat{---}across a diverse set of queries from a variety of sources (e.g., historical Google user queries, dynamically-collected open-ended questions on Reddit, etc.). We find that responses from existing generative search engines are fluent and appear informative, but frequently contain unsupported statements and inaccurate citations: on average, a mere 51.5{\%} of generated sentences are fully supported by citations and only 74.5{\%} of citations support their associated sentence. We believe that these results are concerningly low for systems that may serve as a primary tool for information-seeking users, especially given their facade of trustworthiness. We hope that our results further motivate the development of trustworthy generative search engines and help researchers and users better understand the shortcomings of existing commercial systems. | [
"Liu, Nelson",
"Zhang, Tianyi",
"Liang, Percy"
] | Evaluating Verifiability in Generative Search Engines | findings-emnlp.467 | 2304.09848 | [
"https://github.com/nelson-liu/evaluating-verifiability-in-generative-search-engines"
] | https://huggingface.co/papers/2304.09848 | 0 | 0 | 0 | 3 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.468.bib | https://aclanthology.org/2023.findings-emnlp.468/ | @inproceedings{song-etal-2023-enhancing,
title = "Enhancing Abstractiveness of Summarization Models through Calibrated Distillation",
author = "Song, Hwanjun and
Shalyminov, Igor and
Su, Hang and
Singh, Siffi and
Yao, Kaisheng and
Mansour, Saab",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.468",
doi = "10.18653/v1/2023.findings-emnlp.468",
pages = "7026--7036",
abstract = "In this paper, we propose a novel approach named DisCal to enhance the level of abstractiveness (measured by n-gram overlap) without sacrificing the informativeness (measured by ROUGE) of generated summaries. DisCal exposes diverse pseudo summaries with two supervision to the student model. Firstly, the best pseudo summary is identified in terms of abstractiveness and informativeness and used for sequence-level distillation. Secondly, their ranks are used to ensure the student model to assign higher prediction scores to summaries with higher ranks. Our experiments show that DisCal outperforms prior methods in abstractive summarization distillation, producing highly abstractive and informative summaries.",
}
| In this paper, we propose a novel approach named DisCal to enhance the level of abstractiveness (measured by n-gram overlap) without sacrificing the informativeness (measured by ROUGE) of generated summaries. DisCal exposes diverse pseudo summaries with two supervision to the student model. Firstly, the best pseudo summary is identified in terms of abstractiveness and informativeness and used for sequence-level distillation. Secondly, their ranks are used to ensure the student model to assign higher prediction scores to summaries with higher ranks. Our experiments show that DisCal outperforms prior methods in abstractive summarization distillation, producing highly abstractive and informative summaries. | [
"Song, Hwanjun",
"Shalyminov, Igor",
"Su, Hang",
"Singh, Siffi",
"Yao, Kaisheng",
"Mansour, Saab"
] | Enhancing Abstractiveness of Summarization Models through Calibrated Distillation | findings-emnlp.468 | 2310.13760 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.469.bib | https://aclanthology.org/2023.findings-emnlp.469/ | @inproceedings{ahrens-etal-2023-visually,
title = "Visually Grounded Continual Language Learning with Selective Specialization",
author = "Ahrens, Kyra and
Bengtson, Lennart and
Hee Lee, Jae and
Wermter, Stefan",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.469",
doi = "10.18653/v1/2023.findings-emnlp.469",
pages = "7037--7054",
abstract = "A desirable trait of an artificial agent acting in the visual world is to continually learn a sequence of language-informed tasks while striking a balance between sufficiently specializing in each task and building a generalized knowledge for transfer. Selective specialization, i.e., a careful selection of model components to specialize in each task, is a strategy to provide control over this trade-off. However, the design of selection strategies requires insights on the role of each model component in learning rather specialized or generalizable representations, which poses a gap in current research. Thus, our aim with this work is to provide an extensive analysis of selection strategies for visually grounded continual language learning. Due to the lack of suitable benchmarks for this purpose, we introduce two novel diagnostic datasets that provide enough control and flexibility for a thorough model analysis. We assess various heuristics for module specialization strategies as well as quantifiable measures for two different types of model architectures. Finally, we design conceptually simple approaches based on our analysis that outperform common continual learning baselines. Our results demonstrate the need for further efforts towards better aligning continual learning algorithms with the learning behaviors of individual model parts.",
}
| A desirable trait of an artificial agent acting in the visual world is to continually learn a sequence of language-informed tasks while striking a balance between sufficiently specializing in each task and building a generalized knowledge for transfer. Selective specialization, i.e., a careful selection of model components to specialize in each task, is a strategy to provide control over this trade-off. However, the design of selection strategies requires insights on the role of each model component in learning rather specialized or generalizable representations, which poses a gap in current research. Thus, our aim with this work is to provide an extensive analysis of selection strategies for visually grounded continual language learning. Due to the lack of suitable benchmarks for this purpose, we introduce two novel diagnostic datasets that provide enough control and flexibility for a thorough model analysis. We assess various heuristics for module specialization strategies as well as quantifiable measures for two different types of model architectures. Finally, we design conceptually simple approaches based on our analysis that outperform common continual learning baselines. Our results demonstrate the need for further efforts towards better aligning continual learning algorithms with the learning behaviors of individual model parts. | [
"Ahrens, Kyra",
"Bengtson, Lennart",
"Hee Lee, Jae",
"Wermter, Stefan"
] | Visually Grounded Continual Language Learning with Selective Specialization | findings-emnlp.469 | 2310.15571 | [
"https://github.com/ky-ah/selective-lilac"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.470.bib | https://aclanthology.org/2023.findings-emnlp.470/ | @inproceedings{zhong-etal-2023-romqa,
title = "{R}o{MQA}: A Benchmark for Robust, Multi-evidence, Multi-answer Question Answering",
author = "Zhong, Victor and
Shi, Weijia and
Yih, Wen-tau and
Zettlemoyer, Luke",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.470",
doi = "10.18653/v1/2023.findings-emnlp.470",
pages = "7055--7067",
abstract = "We introduce RoMQA, the first benchmark for robust, multi-evidence, multi-answer question answering (QA). RoMQA contains clusters of questions that are derived from related constraints mined from the Wikidata knowledge graph. RoMQA evaluates robustness of QA models to varying constraints by measuring worst-case performance within each question cluster. Compared to prior QA datasets, RoMQA has more human-written questions that require reasoning over more evidence text and have, on average, many more correct answers. In addition, human annotators rate RoMQA questions as more natural or likely to be asked by people. We evaluate state-of-the-art large language models in zero-shot, few-shot, and fine-tuning settings, and find that RoMQA is challenging: zeroshot and few-shot models perform similarly to naive baselines, while supervised retrieval methods perform well below gold evidence upper bounds. Moreover, existing models are not robust to variations in question constraints, but can be made more robust by tuning on clusters of related questions. Our results show that RoMQA is a challenging benchmark for large language models, and provides a quantifiable test to build more robust QA methods.",
}
| We introduce RoMQA, the first benchmark for robust, multi-evidence, multi-answer question answering (QA). RoMQA contains clusters of questions that are derived from related constraints mined from the Wikidata knowledge graph. RoMQA evaluates robustness of QA models to varying constraints by measuring worst-case performance within each question cluster. Compared to prior QA datasets, RoMQA has more human-written questions that require reasoning over more evidence text and have, on average, many more correct answers. In addition, human annotators rate RoMQA questions as more natural or likely to be asked by people. We evaluate state-of-the-art large language models in zero-shot, few-shot, and fine-tuning settings, and find that RoMQA is challenging: zeroshot and few-shot models perform similarly to naive baselines, while supervised retrieval methods perform well below gold evidence upper bounds. Moreover, existing models are not robust to variations in question constraints, but can be made more robust by tuning on clusters of related questions. Our results show that RoMQA is a challenging benchmark for large language models, and provides a quantifiable test to build more robust QA methods. | [
"Zhong, Victor",
"Shi, Weijia",
"Yih, Wen-tau",
"Zettlemoyer, Luke"
] | RoMQA: A Benchmark for Robust, Multi-evidence, Multi-answer Question Answering | findings-emnlp.470 | 2210.14353 | [
"https://github.com/facebookresearch/romqa"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.471.bib | https://aclanthology.org/2023.findings-emnlp.471/ | @inproceedings{wei-etal-2023-leveraging,
title = "Leveraging Multiple Teachers for Test-Time Adaptation of Language-Guided Classifiers",
author = "Wei, Kangda and
Ghosh, Sayan and
Menon, Rakesh and
Srivastava, Shashank",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.471",
doi = "10.18653/v1/2023.findings-emnlp.471",
pages = "7068--7088",
abstract = "Recent approaches have explored language- guided classifiers capable of classifying examples from novel tasks when provided with task-specific natural language explanations, instructions or prompts (Sanh et al., 2022; R. Menon et al., 2022). While these classifiers can generalize in zero-shot settings, their task performance often varies substantially between different language explanations in unpredictable ways (Lu et al., 2022; Gonen et al., 2022). Also, current approaches fail to leverage unlabeled examples that may be available in many scenarios. Here, we introduce TALC, a framework that uses data programming to adapt a language-guided classifier for a new task during inference when provided with explanations from multiple teachers and unlabeled test examples. Our results show that TALC consistently outperforms a competitive baseline from prior work by an impressive 9.3{\%} (relative improvement). Further, we demonstrate the robustness of TALC to variations in the quality and quantity of provided explanations, highlighting its potential in scenarios where learning from multiple teachers or a crowd is involved. Our code is available at: https://github.com/WeiKangda/TALC.git.",
}
| Recent approaches have explored language- guided classifiers capable of classifying examples from novel tasks when provided with task-specific natural language explanations, instructions or prompts (Sanh et al., 2022; R. Menon et al., 2022). While these classifiers can generalize in zero-shot settings, their task performance often varies substantially between different language explanations in unpredictable ways (Lu et al., 2022; Gonen et al., 2022). Also, current approaches fail to leverage unlabeled examples that may be available in many scenarios. Here, we introduce TALC, a framework that uses data programming to adapt a language-guided classifier for a new task during inference when provided with explanations from multiple teachers and unlabeled test examples. Our results show that TALC consistently outperforms a competitive baseline from prior work by an impressive 9.3{\%} (relative improvement). Further, we demonstrate the robustness of TALC to variations in the quality and quantity of provided explanations, highlighting its potential in scenarios where learning from multiple teachers or a crowd is involved. Our code is available at: https://github.com/WeiKangda/TALC.git. | [
"Wei, Kangda",
"Ghosh, Sayan",
"Menon, Rakesh",
"Srivastava, Shashank"
] | Leveraging Multiple Teachers for Test-Time Adaptation of Language-Guided Classifiers | findings-emnlp.471 | 2311.07538 | [
"https://github.com/weikangda/talc"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.472.bib | https://aclanthology.org/2023.findings-emnlp.472/ | @inproceedings{li-etal-2023-summarizing,
title = "Summarizing Multiple Documents with Conversational Structure for Meta-Review Generation",
author = "Li, Miao and
Hovy, Eduard and
Lau, Jey",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.472",
doi = "10.18653/v1/2023.findings-emnlp.472",
pages = "7089--7112",
abstract = "We present PeerSum, a novel dataset for generating meta-reviews of scientific papers. The meta-reviews can be interpreted as abstractive summaries of reviews, multi-turn discussions and the paper abstract. These source documents have a rich inter-document relationship with an explicit hierarchical conversational structure, cross-references and (occasionally) conflicting information. To introduce the structural inductive bias into pre-trained language models, we introduce RAMMER (Relationship-aware Multi-task Meta-review Generator), a model that uses sparse attention based on the conversational structure and a multi-task training objective that predicts metadata features (e.g., review ratings). Our experimental results show that RAMMER outperforms other strong baseline models in terms of a suite of automatic evaluation metrics. Further analyses, however, reveal that RAMMER and other models struggle to handle conflicts in source documents, suggesting meta-review generation is a challenging task and a promising avenue for further research.",
}
| We present PeerSum, a novel dataset for generating meta-reviews of scientific papers. The meta-reviews can be interpreted as abstractive summaries of reviews, multi-turn discussions and the paper abstract. These source documents have a rich inter-document relationship with an explicit hierarchical conversational structure, cross-references and (occasionally) conflicting information. To introduce the structural inductive bias into pre-trained language models, we introduce RAMMER (Relationship-aware Multi-task Meta-review Generator), a model that uses sparse attention based on the conversational structure and a multi-task training objective that predicts metadata features (e.g., review ratings). Our experimental results show that RAMMER outperforms other strong baseline models in terms of a suite of automatic evaluation metrics. Further analyses, however, reveal that RAMMER and other models struggle to handle conflicts in source documents, suggesting meta-review generation is a challenging task and a promising avenue for further research. | [
"Li, Miao",
"Hovy, Eduard",
"Lau, Jey"
] | Summarizing Multiple Documents with Conversational Structure for Meta-Review Generation | findings-emnlp.472 | 2305.01498 | [
"https://github.com/oaimli/peersum"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.473.bib | https://aclanthology.org/2023.findings-emnlp.473/ | @inproceedings{singh-etal-2023-viphy,
title = "{VIPHY}: Probing {``}Visible{''} Physical Commonsense Knowledge",
author = "Singh, Shikhar and
Qasemi, Ehsan and
Chen, Muhao",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.473",
doi = "10.18653/v1/2023.findings-emnlp.473",
pages = "7113--7128",
abstract = "Vision-language models (VLMs) have shown remarkable performance on visual reasoning tasks (e.g. attributes, location). While such tasks measure the requisite knowledge to ground and reason over a given visual instance, they do not, however, measure the ability of VLMs to retain and generalize such knowledge. In this work, we evaluate VLMs{'} ability to acquire {``}visible{''} physical knowledge {--} the information that is easily accessible from images of static scenes, particularly along the dimensions of object color, size, and space. We build an automatic pipeline to derive a comprehensive knowledge resource for calibrating and probing these models. Our results indicate a severe gap between model and human performance across all three dimensions. Furthermore, we demonstrate that a caption pretrained LM significantly outperforms VLMs on both size and spatial tasks {--} highlighting that despite sufficient access to ground language with visual modality, they struggle to retain such knowledge.",
}
| Vision-language models (VLMs) have shown remarkable performance on visual reasoning tasks (e.g. attributes, location). While such tasks measure the requisite knowledge to ground and reason over a given visual instance, they do not, however, measure the ability of VLMs to retain and generalize such knowledge. In this work, we evaluate VLMs{'} ability to acquire {``}visible{''} physical knowledge {--} the information that is easily accessible from images of static scenes, particularly along the dimensions of object color, size, and space. We build an automatic pipeline to derive a comprehensive knowledge resource for calibrating and probing these models. Our results indicate a severe gap between model and human performance across all three dimensions. Furthermore, we demonstrate that a caption pretrained LM significantly outperforms VLMs on both size and spatial tasks {--} highlighting that despite sufficient access to ground language with visual modality, they struggle to retain such knowledge. | [
"Singh, Shikhar",
"Qasemi, Ehsan",
"Chen, Muhao"
] | VIPHY: Probing “Visible” Physical Commonsense Knowledge | findings-emnlp.473 | [
"https://github.com/axe--/viphy"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.474.bib | https://aclanthology.org/2023.findings-emnlp.474/ | @inproceedings{li-etal-2023-two,
title = "Two Directions for Clinical Data Generation with Large Language Models: Data-to-Label and Label-to-Data",
author = "Li, Rumeng and
Wang, Xun and
Yu, Hong",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.474",
doi = "10.18653/v1/2023.findings-emnlp.474",
pages = "7129--7143",
abstract = "Large language models (LLMs) can generate natural language texts for various domains and tasks, but their potential for clinical text mining, a domain with scarce, sensitive, and imbalanced medical data, is under-explored. We investigate whether LLMs can augment clinical data for detecting Alzheimer{'}s Disease (AD)-related signs and symptoms from electronic health records (EHRs), a challenging task that requires high expertise. We create a novel pragmatic taxonomy for AD sign and symptom progression based on expert knowledge and generated three datasets: (1) a gold dataset annotated by human experts on longitudinal EHRs of AD patients; (2) a silver dataset created by the data-to-label method, which labels sentences from a public EHR collection with AD-related signs and symptoms; and (3) a bronze dataset created by the label-to-data method which generates sentences with AD-related signs and symptoms based on the label definition. We train a system to detect AD-related signs and symptoms from EHRs. We find that the silver and bronze datasets improves the system performance, outperforming the system using only the gold dataset. This shows that LLMs can generate synthetic clinical data for a complex task by incorporating expert knowledge, and our label-to-data method can produce datasets that are free of sensitive information, while maintaining acceptable quality.",
}
| Large language models (LLMs) can generate natural language texts for various domains and tasks, but their potential for clinical text mining, a domain with scarce, sensitive, and imbalanced medical data, is under-explored. We investigate whether LLMs can augment clinical data for detecting Alzheimer{'}s Disease (AD)-related signs and symptoms from electronic health records (EHRs), a challenging task that requires high expertise. We create a novel pragmatic taxonomy for AD sign and symptom progression based on expert knowledge and generated three datasets: (1) a gold dataset annotated by human experts on longitudinal EHRs of AD patients; (2) a silver dataset created by the data-to-label method, which labels sentences from a public EHR collection with AD-related signs and symptoms; and (3) a bronze dataset created by the label-to-data method which generates sentences with AD-related signs and symptoms based on the label definition. We train a system to detect AD-related signs and symptoms from EHRs. We find that the silver and bronze datasets improves the system performance, outperforming the system using only the gold dataset. This shows that LLMs can generate synthetic clinical data for a complex task by incorporating expert knowledge, and our label-to-data method can produce datasets that are free of sensitive information, while maintaining acceptable quality. | [
"Li, Rumeng",
"Wang, Xun",
"Yu, Hong"
] | Two Directions for Clinical Data Generation with Large Language Models: Data-to-Label and Label-to-Data | findings-emnlp.474 | 2401.06774 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.475.bib | https://aclanthology.org/2023.findings-emnlp.475/ | @inproceedings{li-etal-2023-stylized,
title = "Stylized Dialogue Generation with Feature-Guided Knowledge Augmentation",
author = "Li, Jinpeng and
Zhang, Zekai and
Chen, Xiuying and
Zhao, Dongyan and
Yan, Rui",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.475",
doi = "10.18653/v1/2023.findings-emnlp.475",
pages = "7144--7157",
abstract = "Stylized dialogue generation systems aim to produce coherent and context-aware dialogues while effectively emulating the desired style. Generating stylized dialogue is valuable yet challenging due to the scarce parallel data. Existing methods often synthesize pseudo data through back translation, yet suffer from noisy and context-agnostic style signals caused by insufficient guidance on target style features. To address this, we propose the knowledge-augmented stylized dialogue generation model, which includes a feature-guided style knowledge selection module that utilizes context and response features. Specifically, we retrieve dialogue-related style sentences from style corpus to explicitly provide clear style signals. We design a feature-guided selection module with response-related contrastive learning and style responsiveness Kullback-Leibler losses to enhance generation at both semantic and stylized levels. Our approach demonstrates satisfactory performance on two public stylized dialogue benchmarks in both automatic and human evaluations.",
}
| Stylized dialogue generation systems aim to produce coherent and context-aware dialogues while effectively emulating the desired style. Generating stylized dialogue is valuable yet challenging due to the scarce parallel data. Existing methods often synthesize pseudo data through back translation, yet suffer from noisy and context-agnostic style signals caused by insufficient guidance on target style features. To address this, we propose the knowledge-augmented stylized dialogue generation model, which includes a feature-guided style knowledge selection module that utilizes context and response features. Specifically, we retrieve dialogue-related style sentences from style corpus to explicitly provide clear style signals. We design a feature-guided selection module with response-related contrastive learning and style responsiveness Kullback-Leibler losses to enhance generation at both semantic and stylized levels. Our approach demonstrates satisfactory performance on two public stylized dialogue benchmarks in both automatic and human evaluations. | [
"Li, Jinpeng",
"Zhang, Zekai",
"Chen, Xiuying",
"Zhao, Dongyan",
"Yan, Rui"
] | Stylized Dialogue Generation with Feature-Guided Knowledge Augmentation | findings-emnlp.475 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.476.bib | https://aclanthology.org/2023.findings-emnlp.476/ | @inproceedings{starace-etal-2023-probing,
title = "Probing {LLM}s for Joint Encoding of Linguistic Categories",
author = "Starace, Giulio and
Papakostas, Konstantinos and
Choenni, Rochelle and
Panagiotopoulos, Apostolos and
Rosati, Matteo and
Leidinger, Alina and
Shutova, Ekaterina",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.476",
doi = "10.18653/v1/2023.findings-emnlp.476",
pages = "7158--7179",
abstract = "Large Language Models (LLMs) exhibit impressive performance on a range of NLP tasks, due to the general-purpose linguistic knowledge acquired during pretraining. Existing model interpretability research (Tenney et al., 2019) suggests that a linguistic hierarchy emerges in the LLM layers, with lower layers better suited to solving syntactic tasks and higher layers employed for semantic processing. Yet, little is known about how encodings of different linguistic phenomena interact within the models and to what extent processing of linguistically-related categories relies on the same, shared model representations. In this paper, we propose a framework for testing the joint encoding of linguistic categories in LLMs. Focusing on syntax, we find evidence of joint encoding both at the same (related part-of-speech (POS) classes) and different (POS classes and related syntactic dependency relations) levels of linguistic hierarchy. Our cross-lingual experiments show that the same patterns hold across languages in multilingual LLMs.",
}
| Large Language Models (LLMs) exhibit impressive performance on a range of NLP tasks, due to the general-purpose linguistic knowledge acquired during pretraining. Existing model interpretability research (Tenney et al., 2019) suggests that a linguistic hierarchy emerges in the LLM layers, with lower layers better suited to solving syntactic tasks and higher layers employed for semantic processing. Yet, little is known about how encodings of different linguistic phenomena interact within the models and to what extent processing of linguistically-related categories relies on the same, shared model representations. In this paper, we propose a framework for testing the joint encoding of linguistic categories in LLMs. Focusing on syntax, we find evidence of joint encoding both at the same (related part-of-speech (POS) classes) and different (POS classes and related syntactic dependency relations) levels of linguistic hierarchy. Our cross-lingual experiments show that the same patterns hold across languages in multilingual LLMs. | [
"Starace, Giulio",
"Papakostas, Konstantinos",
"Choenni, Rochelle",
"Panagiotopoulos, Apostolos",
"Rosati, Matteo",
"Leidinger, Alina",
"Shutova, Ekaterina"
] | Probing LLMs for Joint Encoding of Linguistic Categories | findings-emnlp.476 | 2310.18696 | [
"https://github.com/thesofakillers/infoshare"
] | https://huggingface.co/papers/2310.18696 | 2 | 1 | 0 | 7 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.477.bib | https://aclanthology.org/2023.findings-emnlp.477/ | @inproceedings{neerudu-etal-2023-robustness,
title = "On Robustness of Finetuned Transformer-based {NLP} Models",
author = "Neerudu, Pavan Kalyan Reddy and
Oota, Subba and
Marreddy, Mounika and
Kagita, Venkateswara and
Gupta, Manish",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.477",
doi = "10.18653/v1/2023.findings-emnlp.477",
pages = "7180--7195",
abstract = "Transformer-based pretrained models like BERT, GPT-2 and T5 have been finetuned for a large number of natural language processing (NLP) tasks, and have been shown to be very effective. However, while finetuning, what changes across layers in these models with respect to pretrained checkpoints is under-studied. Further, how robust are these models to perturbations in input text? Does the robustness vary depending on the NLP task for which the models have been finetuned? While there exists some work on studying robustness of BERT finetuned for a few NLP tasks, there is no rigorous study which compares this robustness across encoder only, decoder only and encoder-decoder models. In this paper, we characterize changes between pretrained and finetuned language model representations across layers using two metrics: CKA and STIR. Further, we study the robustness of three language models (BERT, GPT-2 and T5) with eight different text perturbations on classification tasks from General Language Understanding Evaluation (GLUE) benchmark, and generation tasks like summarization, free-form generation and question generation. GPT-2 representations are more robust than BERT and T5 across multiple types of input perturbation. Although models exhibit good robustness broadly, dropping nouns, verbs or changing characters are the most impactful. Overall, this study provides valuable insights into perturbation-specific weaknesses of popular Transformer-based models which should be kept in mind when passing inputs.",
}
| Transformer-based pretrained models like BERT, GPT-2 and T5 have been finetuned for a large number of natural language processing (NLP) tasks, and have been shown to be very effective. However, while finetuning, what changes across layers in these models with respect to pretrained checkpoints is under-studied. Further, how robust are these models to perturbations in input text? Does the robustness vary depending on the NLP task for which the models have been finetuned? While there exists some work on studying robustness of BERT finetuned for a few NLP tasks, there is no rigorous study which compares this robustness across encoder only, decoder only and encoder-decoder models. In this paper, we characterize changes between pretrained and finetuned language model representations across layers using two metrics: CKA and STIR. Further, we study the robustness of three language models (BERT, GPT-2 and T5) with eight different text perturbations on classification tasks from General Language Understanding Evaluation (GLUE) benchmark, and generation tasks like summarization, free-form generation and question generation. GPT-2 representations are more robust than BERT and T5 across multiple types of input perturbation. Although models exhibit good robustness broadly, dropping nouns, verbs or changing characters are the most impactful. Overall, this study provides valuable insights into perturbation-specific weaknesses of popular Transformer-based models which should be kept in mind when passing inputs. | [
"Neerudu, Pavan Kalyan Reddy",
"Oota, Subba",
"Marreddy, Mounika",
"Kagita, Venkateswara",
"Gupta, Manish"
] | On Robustness of Finetuned Transformer-based NLP Models | findings-emnlp.477 | 2305.14453 | [
"https://github.com/pavanneerudu/robustness-of-transformers-models"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.478.bib | https://aclanthology.org/2023.findings-emnlp.478/ | @inproceedings{wang-etal-2023-measuring,
title = "Measuring and Mitigating Constraint Violations of In-Context Learning for Utterance-to-{API} Semantic Parsing",
author = "Wang, Shufan and
Jean, S{\'e}bastien and
Sengupta, Sailik and
Gung, James and
Pappas, Nikolaos and
Zhang, Yi",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.478",
doi = "10.18653/v1/2023.findings-emnlp.478",
pages = "7196--7207",
abstract = "In executable task-oriented semantic parsing, the system aims to translate users{'} utterances in natural language to machine-interpretable programs (API calls) that can be executed according to pre-defined API specifications. With the popularity of Large Language Models (LLMs), in-context learning offers a strong baseline for such scenarios, especially in data-limited regimes. However, LLMs are known to hallucinate and therefore pose a formidable challenge in constraining generated content. Thus, it remains uncertain if LLMs can effectively perform task-oriented utterance-to-API generation, where respecting the API{'}s structural and task-specific constraints is crucial. In this work, we seek to measure, analyze and mitigate such constraints violations. First, we identify the categories of various constraints in obtaining API-semantics from task-oriented utterances, and define fine-grained metrics that complement traditional ones. Second, we leverage these metrics to conduct a detailed error analysis of constraints violations seen in state-of-the-art LLMs, which motivates us to investigate two popular mitigation strategies{--} Semantic-Retrieval of Demonstrations (SRD) and API-aware Constrained Decoding (API-CD). Our experiments show that these strategies are effective at reducing constraints violations and improving the quality of the generated API calls, but require careful consideration given their implementation complexity and latency.",
}
| In executable task-oriented semantic parsing, the system aims to translate users{'} utterances in natural language to machine-interpretable programs (API calls) that can be executed according to pre-defined API specifications. With the popularity of Large Language Models (LLMs), in-context learning offers a strong baseline for such scenarios, especially in data-limited regimes. However, LLMs are known to hallucinate and therefore pose a formidable challenge in constraining generated content. Thus, it remains uncertain if LLMs can effectively perform task-oriented utterance-to-API generation, where respecting the API{'}s structural and task-specific constraints is crucial. In this work, we seek to measure, analyze and mitigate such constraints violations. First, we identify the categories of various constraints in obtaining API-semantics from task-oriented utterances, and define fine-grained metrics that complement traditional ones. Second, we leverage these metrics to conduct a detailed error analysis of constraints violations seen in state-of-the-art LLMs, which motivates us to investigate two popular mitigation strategies{--} Semantic-Retrieval of Demonstrations (SRD) and API-aware Constrained Decoding (API-CD). Our experiments show that these strategies are effective at reducing constraints violations and improving the quality of the generated API calls, but require careful consideration given their implementation complexity and latency. | [
"Wang, Shufan",
"Jean, S{\\'e}bastien",
"Sengupta, Sailik",
"Gung, James",
"Pappas, Nikolaos",
"Zhang, Yi"
] | Measuring and Mitigating Constraint Violations of In-Context Learning for Utterance-to-API Semantic Parsing | findings-emnlp.478 | 2305.15338 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.479.bib | https://aclanthology.org/2023.findings-emnlp.479/ | @inproceedings{primadhanty-quattoni-2023-entity,
title = "Entity Disambiguation on a Tight Labeling Budget",
author = "Primadhanty, Audi and
Quattoni, Ariadna",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.479",
doi = "10.18653/v1/2023.findings-emnlp.479",
pages = "7208--7215",
abstract = "Many real-world NLP applications face the challenge of training an entity disambiguation model for a specific domain with a small labeling budget. In this setting there is often access to a large unlabeled pool of documents. It is then natural to ask the question: which samples should be selected for annotation? In this paper we propose a solution that combines feature diversity with low rank correction. Our sampling strategy is formulated in the context of bilinear tensor models. Our experiments show that the proposed approach can significantly reduce the amount of labeled data necessary to achieve a given performance.",
}
| Many real-world NLP applications face the challenge of training an entity disambiguation model for a specific domain with a small labeling budget. In this setting there is often access to a large unlabeled pool of documents. It is then natural to ask the question: which samples should be selected for annotation? In this paper we propose a solution that combines feature diversity with low rank correction. Our sampling strategy is formulated in the context of bilinear tensor models. Our experiments show that the proposed approach can significantly reduce the amount of labeled data necessary to achieve a given performance. | [
"Primadhanty, Audi",
"Quattoni, Ariadna"
] | Entity Disambiguation on a Tight Labeling Budget | findings-emnlp.479 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.480.bib | https://aclanthology.org/2023.findings-emnlp.480/ | @inproceedings{xiao-etal-2023-topic,
title = "Topic-{DPR}: Topic-based Prompts for Dense Passage Retrieval",
author = "Xiao, Qingfa and
Li, Shuangyin and
Chen, Lei",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.480",
doi = "10.18653/v1/2023.findings-emnlp.480",
pages = "7216--7225",
abstract = "Prompt-based learning{'}s efficacy across numerous natural language processing tasks has led to its integration into dense passage retrieval. Prior research has mainly focused on enhancing the semantic understanding of pre-trained language models by optimizing a single vector as a continuous prompt. This approach, however, leads to a semantic space collapse; identical semantic information seeps into all representations, causing their distributions to converge in a restricted region. This hinders differentiation between relevant and irrelevant passages during dense retrieval. To tackle this issue, we present Topic-DPR, a dense passage retrieval model that uses topic-based prompts. Unlike the single prompt method, multiple topic-based prompts are established over a probabilistic simplex and optimized simultaneously through contrastive learning. This encourages representations to align with their topic distributions, improving space uniformity. Furthermore, we introduce a novel positive and negative sampling strategy, leveraging semi-structured data to boost dense retrieval efficiency. Experimental results from two datasets affirm that our method surpasses previous state-of-the-art retrieval techniques.",
}
| Prompt-based learning{'}s efficacy across numerous natural language processing tasks has led to its integration into dense passage retrieval. Prior research has mainly focused on enhancing the semantic understanding of pre-trained language models by optimizing a single vector as a continuous prompt. This approach, however, leads to a semantic space collapse; identical semantic information seeps into all representations, causing their distributions to converge in a restricted region. This hinders differentiation between relevant and irrelevant passages during dense retrieval. To tackle this issue, we present Topic-DPR, a dense passage retrieval model that uses topic-based prompts. Unlike the single prompt method, multiple topic-based prompts are established over a probabilistic simplex and optimized simultaneously through contrastive learning. This encourages representations to align with their topic distributions, improving space uniformity. Furthermore, we introduce a novel positive and negative sampling strategy, leveraging semi-structured data to boost dense retrieval efficiency. Experimental results from two datasets affirm that our method surpasses previous state-of-the-art retrieval techniques. | [
"Xiao, Qingfa",
"Li, Shuangyin",
"Chen, Lei"
] | Topic-DPR: Topic-based Prompts for Dense Passage Retrieval | findings-emnlp.480 | 2310.06626 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.481.bib | https://aclanthology.org/2023.findings-emnlp.481/ | @inproceedings{kantharuban-etal-2023-quantifying,
title = "Quantifying the Dialect Gap and its Correlates Across Languages",
author = "Kantharuban, Anjali and
Vuli{\'c}, Ivan and
Korhonen, Anna",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.481",
doi = "10.18653/v1/2023.findings-emnlp.481",
pages = "7226--7245",
abstract = "Historically, researchers and consumers have noticed a decrease in quality when applying NLP tools to minority variants of languages (i.e. Puerto Rican Spanish or Swiss German), but studies exploring this have been limited to a select few languages. Additionally, past studies have mainly been conducted in a monolingual context, so cross-linguistic trends have not been identified and tied to external factors. In this work, we conduct a comprehensive evaluation of the most influential, state-of-the-art large language models (LLMs) across two high-use applications, machine translation and automatic speech recognition, to assess their functionality on the regional dialects of several high- and low-resource languages. Additionally, we analyze how the regional dialect gap is correlated with economic, social, and linguistic factors. The impact of training data, including related factors like dataset size and its construction procedure, is shown to be significant but not consistent across models or languages, meaning a one-size-fits-all approach cannot be taken in solving the dialect gap. This work will lay the foundation for furthering the field of dialectal NLP by laying out evident disparities and identifying possible pathways for addressing them through mindful data collection.",
}
| Historically, researchers and consumers have noticed a decrease in quality when applying NLP tools to minority variants of languages (i.e. Puerto Rican Spanish or Swiss German), but studies exploring this have been limited to a select few languages. Additionally, past studies have mainly been conducted in a monolingual context, so cross-linguistic trends have not been identified and tied to external factors. In this work, we conduct a comprehensive evaluation of the most influential, state-of-the-art large language models (LLMs) across two high-use applications, machine translation and automatic speech recognition, to assess their functionality on the regional dialects of several high- and low-resource languages. Additionally, we analyze how the regional dialect gap is correlated with economic, social, and linguistic factors. The impact of training data, including related factors like dataset size and its construction procedure, is shown to be significant but not consistent across models or languages, meaning a one-size-fits-all approach cannot be taken in solving the dialect gap. This work will lay the foundation for furthering the field of dialectal NLP by laying out evident disparities and identifying possible pathways for addressing them through mindful data collection. | [
"Kantharuban, Anjali",
"Vuli{\\'c}, Ivan",
"Korhonen, Anna"
] | Quantifying the Dialect Gap and its Correlates Across Languages | findings-emnlp.481 | 2310.15135 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.482.bib | https://aclanthology.org/2023.findings-emnlp.482/ | @inproceedings{haotian-etal-2023-recal,
title = "{RECAL}: Sample-Relation Guided Confidence Calibration over Tabular Data",
author = "HaoTian, Wang and
Zhang, Zhen and
Hu, Mengting and
Wang, Qichao and
Chen, Liang and
Bian, Yatao and
Wu, Bingzhe",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.482",
doi = "10.18653/v1/2023.findings-emnlp.482",
pages = "7246--7257",
abstract = "Tabular-format data is widely adopted in various real-world applications. Various machine learning models have achieved remarkable success in both industrial applications and data-science competitions. Despite these successes, most current machine learning methods for tabular data lack accurate confidence estimation, which is needed by some high-risk sensitive applications such as credit modeling and financial fraud detection. In this paper, we study the confidence estimation of machine learning models applied to tabular data. The key finding of our paper is that a real-world tabular dataset typically contains implicit sample relations, and this can further help to obtain a more accurate estimation. To this end, we introduce a general post-training confidence calibration framework named RECAL to calibrate the predictive confidence of current machine learning models by employing graph neural networks to model the relations between different samples. We perform extensive experiments on tabular datasets with both implicit and explicit graph structures and show that RECAL can significantly improve the calibration quality compared to the conventional method without considering the sample relations.",
}
| Tabular-format data is widely adopted in various real-world applications. Various machine learning models have achieved remarkable success in both industrial applications and data-science competitions. Despite these successes, most current machine learning methods for tabular data lack accurate confidence estimation, which is needed by some high-risk sensitive applications such as credit modeling and financial fraud detection. In this paper, we study the confidence estimation of machine learning models applied to tabular data. The key finding of our paper is that a real-world tabular dataset typically contains implicit sample relations, and this can further help to obtain a more accurate estimation. To this end, we introduce a general post-training confidence calibration framework named RECAL to calibrate the predictive confidence of current machine learning models by employing graph neural networks to model the relations between different samples. We perform extensive experiments on tabular datasets with both implicit and explicit graph structures and show that RECAL can significantly improve the calibration quality compared to the conventional method without considering the sample relations. | [
"HaoTian, Wang",
"Zhang, Zhen",
"Hu, Mengting",
"Wang, Qichao",
"Chen, Liang",
"Bian, Yatao",
"Wu, Bingzhe"
] | RECAL: Sample-Relation Guided Confidence Calibration over Tabular Data | findings-emnlp.482 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.483.bib | https://aclanthology.org/2023.findings-emnlp.483/ | @inproceedings{zhang-etal-2023-parameter,
title = "Parameter-Efficient Cross-lingual Transfer of Vision and Language Models via Translation-based Alignment",
author = "Zhang, Zhen and
Wang, Jialu and
Wang, Xin",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.483",
doi = "10.18653/v1/2023.findings-emnlp.483",
pages = "7258--7268",
abstract = "Pre-trained vision and language models such as CLIP have witnessed remarkable success in connecting images and texts with a primary focus on English texts. Despite recent efforts to extend CLIP to support other languages, disparities in performance among different languages have been observed due to uneven resource availability. Additionally, current cross-lingual transfer methods of those pre-trained models would consume excessive resources for a large number of languages. Therefore, we propose a new parameter-efficient cross-lingual transfer learning framework that utilizes a translation-based alignment method to mitigate multilingual disparities and explores parameter-efficient fine-tuning methods for parameter-efficient cross-lingual transfer. Extensive experiments on XTD and Multi30K datasets, covering 11 languages under zero-shot, few-shot, and full-dataset learning scenarios, show that our framework significantly reduces the multilingual disparities among languages and improves cross-lingual transfer results, especially in low-resource scenarios, while only keeping and fine-tuning an extremely small number of parameters compared to the full model (e.g., Our framework only requires 0.16{\%} additional parameters of a full-model for each language in the few-shot learning scenario).",
}
| Pre-trained vision and language models such as CLIP have witnessed remarkable success in connecting images and texts with a primary focus on English texts. Despite recent efforts to extend CLIP to support other languages, disparities in performance among different languages have been observed due to uneven resource availability. Additionally, current cross-lingual transfer methods of those pre-trained models would consume excessive resources for a large number of languages. Therefore, we propose a new parameter-efficient cross-lingual transfer learning framework that utilizes a translation-based alignment method to mitigate multilingual disparities and explores parameter-efficient fine-tuning methods for parameter-efficient cross-lingual transfer. Extensive experiments on XTD and Multi30K datasets, covering 11 languages under zero-shot, few-shot, and full-dataset learning scenarios, show that our framework significantly reduces the multilingual disparities among languages and improves cross-lingual transfer results, especially in low-resource scenarios, while only keeping and fine-tuning an extremely small number of parameters compared to the full model (e.g., Our framework only requires 0.16{\%} additional parameters of a full-model for each language in the few-shot learning scenario). | [
"Zhang, Zhen",
"Wang, Jialu",
"Wang, Xin"
] | Parameter-Efficient Cross-lingual Transfer of Vision and Language Models via Translation-based Alignment | findings-emnlp.483 | 2305.03510 | [
"https://github.com/eric-ai-lab/pectvlm"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.484.bib | https://aclanthology.org/2023.findings-emnlp.484/ | @inproceedings{choubey-etal-2023-lexical,
title = "Lexical Repetitions Lead to Rote Learning: Unveiling the Impact of Lexical Overlap in Train and Test Reference Summaries",
author = "Choubey, Prafulla and
Fabbri, Alexander and
Xiong, Caiming and
Wu, Chien-Sheng",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.484",
doi = "10.18653/v1/2023.findings-emnlp.484",
pages = "7269--7283",
abstract = "Ideal summarization models should generalize to novel summary-worthy content without remembering reference training summaries by rote. However, a single average performance score on the entire test set is inadequate in determining such model competencies. We propose a fine-grained evaluation protocol by partitioning a test set based on the lexical similarity of reference test summaries with training summaries. We observe up to a 5x (1.2x) difference in ROUGE-2 (entity recall) scores between the subsets with the lowest and highest similarity. Next, we show that such training repetitions also make a model vulnerable to rote learning, reproducing data artifacts such as factual errors, especially when reference test summaries are lexically close to training summaries. Consequently, we propose to limit lexical repetitions in training summaries during both supervised fine-tuning and likelihood calibration stages to improve the performance on novel test cases while retaining average performance. Our automatic and human evaluations on novel test subsets and recent news articles show that limiting lexical repetitions in training summaries can prevent rote learning and improve generalization.",
}
| Ideal summarization models should generalize to novel summary-worthy content without remembering reference training summaries by rote. However, a single average performance score on the entire test set is inadequate in determining such model competencies. We propose a fine-grained evaluation protocol by partitioning a test set based on the lexical similarity of reference test summaries with training summaries. We observe up to a 5x (1.2x) difference in ROUGE-2 (entity recall) scores between the subsets with the lowest and highest similarity. Next, we show that such training repetitions also make a model vulnerable to rote learning, reproducing data artifacts such as factual errors, especially when reference test summaries are lexically close to training summaries. Consequently, we propose to limit lexical repetitions in training summaries during both supervised fine-tuning and likelihood calibration stages to improve the performance on novel test cases while retaining average performance. Our automatic and human evaluations on novel test subsets and recent news articles show that limiting lexical repetitions in training summaries can prevent rote learning and improve generalization. | [
"Choubey, Prafulla",
"Fabbri, Alex",
"er",
"Xiong, Caiming",
"Wu, Chien-Sheng"
] | Lexical Repetitions Lead to Rote Learning: Unveiling the Impact of Lexical Overlap in Train and Test Reference Summaries | findings-emnlp.484 | 2311.09458 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.485.bib | https://aclanthology.org/2023.findings-emnlp.485/ | @inproceedings{murty-etal-2023-pseudointelligence,
title = "Pseudointelligence: A Unifying Lens on Language Model Evaluation",
author = "Murty, Shikhar and
Paradise, Orr and
Sharma, Pratyusha",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.485",
doi = "10.18653/v1/2023.findings-emnlp.485",
pages = "7284--7290",
abstract = "With large language models surpassing human performance on an increasing number of benchmarks, we must take a principled approach for targeted evaluation of model capabilities. Inspired by pseudorandomness, we propose pseudointelligence, which captures the maxim that {``}(perceived) intelligence lies in the eye of the beholder.{''} That is, that claims of intelligence are meaningful only when their evaluator is taken into account. Concretely, we propose a complexity-theoretic framework of model evaluation cast as a dynamic interaction between a model and a learned evaluator. We demonstrate that this framework can be used to reason about two case studies in language model evaluation, as well as analyze existing evaluation methods.",
}
| With large language models surpassing human performance on an increasing number of benchmarks, we must take a principled approach for targeted evaluation of model capabilities. Inspired by pseudorandomness, we propose pseudointelligence, which captures the maxim that {``}(perceived) intelligence lies in the eye of the beholder.{''} That is, that claims of intelligence are meaningful only when their evaluator is taken into account. Concretely, we propose a complexity-theoretic framework of model evaluation cast as a dynamic interaction between a model and a learned evaluator. We demonstrate that this framework can be used to reason about two case studies in language model evaluation, as well as analyze existing evaluation methods. | [
"Murty, Shikhar",
"Paradise, Orr",
"Sharma, Pratyusha"
] | Pseudointelligence: A Unifying Lens on Language Model Evaluation | findings-emnlp.485 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.486.bib | https://aclanthology.org/2023.findings-emnlp.486/ | @inproceedings{hahn-etal-2023-gda,
title = "{GDA}: Grammar-based Data Augmentation for Text Classification using Slot Information",
author = "Hahn, Joonghyuk and
Cheon, Hyunjoon and
Orwig, Elizabeth and
Kim, Su-Hyeon and
Ko, Sang-Ki and
Han, Yo-Sub",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.486",
doi = "10.18653/v1/2023.findings-emnlp.486",
pages = "7291--7306",
abstract = "Recent studies propose various data augmentation approaches to resolve the low-resource problem in natural language processing tasks. Data augmentation is a successful solution to this problem and recent strategies give variation on sentence structures to boost performance. However, these approaches can potentially lead to semantic errors and produce semantically noisy data due to the unregulated variation of sentence structures. In an effort to combat these semantic errors, we leverage slot information, the representation of the context of keywords from a sentence, and form a data augmentation strategy which we propose, called GDA. Our strategy employs algorithms that construct and manipulate rules of context-aware grammar, utilizing this slot information. The algorithms extract recurrent patterns by distinguishing words with slots and form the {``}rules of grammar{''}{---}a set of injective relations between a sentence{'}s semantics and its syntactical structure{---}to augment the dataset. The augmentation is done in an automated manner with the constructed rules and thus, GDA is explainable and reliable without any human intervention. We evaluate GDA with state-of-the-art data augmentation techniques, including those using pre-trained language models, and the result illustrates that GDA outperforms all other data augmentation methods by 19.38{\%}. Extensive experiments show that GDA is an effective data augmentation strategy that incorporates word semantics for more accurate and diverse data.",
}
| Recent studies propose various data augmentation approaches to resolve the low-resource problem in natural language processing tasks. Data augmentation is a successful solution to this problem and recent strategies give variation on sentence structures to boost performance. However, these approaches can potentially lead to semantic errors and produce semantically noisy data due to the unregulated variation of sentence structures. In an effort to combat these semantic errors, we leverage slot information, the representation of the context of keywords from a sentence, and form a data augmentation strategy which we propose, called GDA. Our strategy employs algorithms that construct and manipulate rules of context-aware grammar, utilizing this slot information. The algorithms extract recurrent patterns by distinguishing words with slots and form the {``}rules of grammar{''}{---}a set of injective relations between a sentence{'}s semantics and its syntactical structure{---}to augment the dataset. The augmentation is done in an automated manner with the constructed rules and thus, GDA is explainable and reliable without any human intervention. We evaluate GDA with state-of-the-art data augmentation techniques, including those using pre-trained language models, and the result illustrates that GDA outperforms all other data augmentation methods by 19.38{\%}. Extensive experiments show that GDA is an effective data augmentation strategy that incorporates word semantics for more accurate and diverse data. | [
"Hahn, Joonghyuk",
"Cheon, Hyunjoon",
"Orwig, Elizabeth",
"Kim, Su-Hyeon",
"Ko, Sang-Ki",
"Han, Yo-Sub"
] | GDA: Grammar-based Data Augmentation for Text Classification using Slot Information | findings-emnlp.486 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.487.bib | https://aclanthology.org/2023.findings-emnlp.487/ | @inproceedings{oka-hirao-2023-implicit,
title = "Implicit Sense-labeled Connective Recognition as Text Generation",
author = "Oka, Yui and
Hirao, Tsutomu",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.487",
doi = "10.18653/v1/2023.findings-emnlp.487",
pages = "7307--7313",
abstract = "Implicit Discourse Relation Recognition (IDRR) involves identifying the sense label of an implicit connective between adjacent text spans. This has traditionally been approached as a classification task. However, some downstream tasks require more than just a sense label as well as the specific connective used. This paper presents Implicit Sense-labeled Connective Recognition (ISCR), which identifies the implicit connectives and their sense labels between adjacent text spans. ISCR can be treated as a classification task, but a large number of potential categories, sense labels, and uneven distribution of instances among them make this difficult. Instead, this paper handles the task as a text-generation task, using an encoder-decoder model to generate both connectives and their sense labels. Here, we explore a classification method and three kinds of text-generation methods. From our evaluation results on PDTB-3.0, we found that our method outperforms the conventional classification-based method.",
}
| Implicit Discourse Relation Recognition (IDRR) involves identifying the sense label of an implicit connective between adjacent text spans. This has traditionally been approached as a classification task. However, some downstream tasks require more than just a sense label as well as the specific connective used. This paper presents Implicit Sense-labeled Connective Recognition (ISCR), which identifies the implicit connectives and their sense labels between adjacent text spans. ISCR can be treated as a classification task, but a large number of potential categories, sense labels, and uneven distribution of instances among them make this difficult. Instead, this paper handles the task as a text-generation task, using an encoder-decoder model to generate both connectives and their sense labels. Here, we explore a classification method and three kinds of text-generation methods. From our evaluation results on PDTB-3.0, we found that our method outperforms the conventional classification-based method. | [
"Oka, Yui",
"Hirao, Tsutomu"
] | Implicit Sense-labeled Connective Recognition as Text Generation | findings-emnlp.487 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.488.bib | https://aclanthology.org/2023.findings-emnlp.488/ | @inproceedings{lee-etal-2023-vista,
title = "{VISTA}: Visual-Textual Knowledge Graph Representation Learning",
author = "Lee, Jaejun and
Chung, Chanyoung and
Lee, Hochang and
Jo, Sungho and
Whang, Joyce",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.488",
doi = "10.18653/v1/2023.findings-emnlp.488",
pages = "7314--7328",
abstract = "Knowledge graphs represent human knowledge using triplets composed of entities and relations. While most existing knowledge graph embedding methods only consider the structure of a knowledge graph, a few recently proposed multimodal methods utilize images or text descriptions of entities in a knowledge graph. In this paper, we propose visual-textual knowledge graphs (VTKGs), where not only entities but also triplets can be explained using images, and both entities and relations can accompany text descriptions. By compiling visually expressible commonsense knowledge, we construct new benchmark datasets where triplets themselves are explained by images, and the meanings of entities and relations are described using text. We propose VISTA, a knowledge graph representation learning method for VTKGs, which incorporates the visual and textual representations of entities and relations using entity encoding, relation encoding, and triplet decoding transformers. Experiments show that VISTA outperforms state-of-the-art knowledge graph completion methods in real-world VTKGs.",
}
| Knowledge graphs represent human knowledge using triplets composed of entities and relations. While most existing knowledge graph embedding methods only consider the structure of a knowledge graph, a few recently proposed multimodal methods utilize images or text descriptions of entities in a knowledge graph. In this paper, we propose visual-textual knowledge graphs (VTKGs), where not only entities but also triplets can be explained using images, and both entities and relations can accompany text descriptions. By compiling visually expressible commonsense knowledge, we construct new benchmark datasets where triplets themselves are explained by images, and the meanings of entities and relations are described using text. We propose VISTA, a knowledge graph representation learning method for VTKGs, which incorporates the visual and textual representations of entities and relations using entity encoding, relation encoding, and triplet decoding transformers. Experiments show that VISTA outperforms state-of-the-art knowledge graph completion methods in real-world VTKGs. | [
"Lee, Jaejun",
"Chung, Chanyoung",
"Lee, Hochang",
"Jo, Sungho",
"Whang, Joyce"
] | VISTA: Visual-Textual Knowledge Graph Representation Learning | findings-emnlp.488 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.489.bib | https://aclanthology.org/2023.findings-emnlp.489/ | @inproceedings{yang-etal-2023-dynamic,
title = "Dynamic Stashing Quantization for Efficient Transformer Training",
author = "Yang, Guo and
Lo, Daniel and
Mullins, Robert and
Zhao, Yiren",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.489",
doi = "10.18653/v1/2023.findings-emnlp.489",
pages = "7329--7336",
abstract = "Large Language Models (LLMs) have demonstrated impressive performance on a range of Natural Language Processing (NLP) tasks. Unfortunately, the immense amount of computations and memory accesses required for LLM training makes them prohibitively expensive in terms of hardware cost, and thus challenging to deploy in use cases such as on-device learning. In this paper, motivated by the observation that LLM training is memory-bound, we propose a novel dynamic quantization strategy, termed Dynamic Stashing Quantization (DSQ), that puts a special focus on reducing the memory operations, but also enjoys the other benefits of low precision training, such as the reduced arithmetic cost. We conduct a thorough study on two translation tasks (trained-from-scratch) and three classification tasks (fine-tuning). DSQ reduces the amount of arithmetic operations by $20.95\times$ and the number of DRAM operations by $2.55\times$ on IWSLT17 compared to the standard 16-bit fixed-point, which is widely used in on-device learning.",
}
| Large Language Models (LLMs) have demonstrated impressive performance on a range of Natural Language Processing (NLP) tasks. Unfortunately, the immense amount of computations and memory accesses required for LLM training makes them prohibitively expensive in terms of hardware cost, and thus challenging to deploy in use cases such as on-device learning. In this paper, motivated by the observation that LLM training is memory-bound, we propose a novel dynamic quantization strategy, termed Dynamic Stashing Quantization (DSQ), that puts a special focus on reducing the memory operations, but also enjoys the other benefits of low precision training, such as the reduced arithmetic cost. We conduct a thorough study on two translation tasks (trained-from-scratch) and three classification tasks (fine-tuning). DSQ reduces the amount of arithmetic operations by $20.95\times$ and the number of DRAM operations by $2.55\times$ on IWSLT17 compared to the standard 16-bit fixed-point, which is widely used in on-device learning. | [
"Yang, Guo",
"Lo, Daniel",
"Mullins, Robert",
"Zhao, Yiren"
] | Dynamic Stashing Quantization for Efficient Transformer Training | findings-emnlp.489 | 2303.05295 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.490.bib | https://aclanthology.org/2023.findings-emnlp.490/ | @inproceedings{shui-etal-2023-comprehensive,
title = "A Comprehensive Evaluation of Large Language Models on Legal Judgment Prediction",
author = "Shui, Ruihao and
Cao, Yixin and
Wang, Xiang and
Chua, Tat-Seng",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.490",
doi = "10.18653/v1/2023.findings-emnlp.490",
pages = "7337--7348",
abstract = "Large language models (LLMs) have demonstrated great potential for domain-specific applications, such as the law domain. However, recent disputes over GPT-4{'}s law evaluation raise questions concerning their performance in real-world legal tasks. To systematically investigate their competency in the law, we design practical baseline solutions based on LLMs and test on the task of legal judgment prediction. In our solutions, LLMs can work alone to answer open questions or coordinate with an information retrieval (IR) system to learn from similar cases or solve simplified multi-choice questions. We show that similar cases and multi-choice options, namely label candidates, included in prompts can help LLMs recall domain knowledge that is critical for expertise legal reasoning. We additionally present an intriguing paradox wherein an IR system surpasses the performance of LLM+IR due to limited gains acquired by weaker LLMs from powerful IR systems. In such case, the role of LLMs becomes redundant. Our evaluation pipeline can be easily extended into other tasks to facilitate evaluations in other domains. Code is available at https://github.com/srhthu/LM-CompEval-Legal",
}
| Large language models (LLMs) have demonstrated great potential for domain-specific applications, such as the law domain. However, recent disputes over GPT-4{'}s law evaluation raise questions concerning their performance in real-world legal tasks. To systematically investigate their competency in the law, we design practical baseline solutions based on LLMs and test on the task of legal judgment prediction. In our solutions, LLMs can work alone to answer open questions or coordinate with an information retrieval (IR) system to learn from similar cases or solve simplified multi-choice questions. We show that similar cases and multi-choice options, namely label candidates, included in prompts can help LLMs recall domain knowledge that is critical for expertise legal reasoning. We additionally present an intriguing paradox wherein an IR system surpasses the performance of LLM+IR due to limited gains acquired by weaker LLMs from powerful IR systems. In such case, the role of LLMs becomes redundant. Our evaluation pipeline can be easily extended into other tasks to facilitate evaluations in other domains. Code is available at https://github.com/srhthu/LM-CompEval-Legal | [
"Shui, Ruihao",
"Cao, Yixin",
"Wang, Xiang",
"Chua, Tat-Seng"
] | A Comprehensive Evaluation of Large Language Models on Legal Judgment Prediction | findings-emnlp.490 | 2310.11761 | [
"https://github.com/srhthu/lm-compeval-legal"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.491.bib | https://aclanthology.org/2023.findings-emnlp.491/ | @inproceedings{gautam-etal-2023-lightweight,
title = "A Lightweight Method to Generate Unanswerable Questions in {E}nglish",
author = "Gautam, Vagrant and
Zhang, Miaoran and
Klakow, Dietrich",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.491",
doi = "10.18653/v1/2023.findings-emnlp.491",
pages = "7349--7360",
abstract = "If a question cannot be answered with the available information, robust systems for question answering (QA) should know *not* to answer. One way to build QA models that do this is with additional training data comprised of unanswerable questions, created either by employing annotators or through automated methods for unanswerable question generation. To show that the model complexity of existing automated approaches is not justified, we examine a simpler data augmentation method for unanswerable question generation in English: performing antonym and entity swaps on answerable questions. Compared to the prior state-of-the-art, data generated with our training-free and lightweight strategy results in better models (+1.6 F1 points on SQuAD 2.0 data with BERT-large), and has higher human-judged relatedness and readability. We quantify the raw benefits of our approach compared to no augmentation across multiple encoder models, using different amounts of generated data, and also on TydiQA-MinSpan data (+9.3 F1 points with BERT-large). Our results establish swaps as a simple but strong baseline for future work.",
}
| If a question cannot be answered with the available information, robust systems for question answering (QA) should know *not* to answer. One way to build QA models that do this is with additional training data comprised of unanswerable questions, created either by employing annotators or through automated methods for unanswerable question generation. To show that the model complexity of existing automated approaches is not justified, we examine a simpler data augmentation method for unanswerable question generation in English: performing antonym and entity swaps on answerable questions. Compared to the prior state-of-the-art, data generated with our training-free and lightweight strategy results in better models (+1.6 F1 points on SQuAD 2.0 data with BERT-large), and has higher human-judged relatedness and readability. We quantify the raw benefits of our approach compared to no augmentation across multiple encoder models, using different amounts of generated data, and also on TydiQA-MinSpan data (+9.3 F1 points with BERT-large). Our results establish swaps as a simple but strong baseline for future work. | [
"Gautam, Vagrant",
"Zhang, Miaoran",
"Klakow, Dietrich"
] | A Lightweight Method to Generate Unanswerable Questions in English | findings-emnlp.491 | 2310.19403 | [
"https://github.com/uds-lsv/unanswerable-question-generation"
] | https://huggingface.co/papers/2310.19403 | 1 | 0 | 0 | 3 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.492.bib | https://aclanthology.org/2023.findings-emnlp.492/ | @inproceedings{chen-etal-2023-automatic,
title = "Automatic Evaluate Dialogue Appropriateness by Using Dialogue Act",
author = "Chen, Bao and
Wang, Yuanjie and
Liu, Zeming and
Guo, Yuhang",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.492",
doi = "10.18653/v1/2023.findings-emnlp.492",
pages = "7361--7372",
abstract = "Evaluation of dialogue systems requires assessing various aspects, among which appropriateness holds significance as a core element of communicative language competence. However, current evaluations heavily rely on human judgments, which are time-consuming, labor-intensive, prone to biases, and lacking objectivity. In this paper, we introduce Dialogue Act Appropriateness (DAA), a novel method that utilizes the underlying patterns of dialogue act transitions to evaluate the appropriateness of chatbot responses. We learn transition patterns from human-human dialogue corpora, evaluating chatbot appropriateness by measuring the similarity of their transition patterns to those observed in human-human dialogues. To validate DAA, we annotate a test dataset by manually evaluating the appropriateness of dialogues from multiple chatbot systems. The experimental results demonstrate a strong correlation between our evaluation metric and human ratings, establishing the reliability of DAA as a measure of dialogue appropriateness.",
}
| Evaluation of dialogue systems requires assessing various aspects, among which appropriateness holds significance as a core element of communicative language competence. However, current evaluations heavily rely on human judgments, which are time-consuming, labor-intensive, prone to biases, and lacking objectivity. In this paper, we introduce Dialogue Act Appropriateness (DAA), a novel method that utilizes the underlying patterns of dialogue act transitions to evaluate the appropriateness of chatbot responses. We learn transition patterns from human-human dialogue corpora, evaluating chatbot appropriateness by measuring the similarity of their transition patterns to those observed in human-human dialogues. To validate DAA, we annotate a test dataset by manually evaluating the appropriateness of dialogues from multiple chatbot systems. The experimental results demonstrate a strong correlation between our evaluation metric and human ratings, establishing the reliability of DAA as a measure of dialogue appropriateness. | [
"Chen, Bao",
"Wang, Yuanjie",
"Liu, Zeming",
"Guo, Yuhang"
] | Automatic Evaluate Dialogue Appropriateness by Using Dialogue Act | findings-emnlp.492 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.493.bib | https://aclanthology.org/2023.findings-emnlp.493/ | @inproceedings{jin-etal-2023-tabprompt,
title = "{T}ab{P}rompt: Graph-based Pre-training and Prompting for Few-shot Table Understanding",
author = "Jin, Rihui and
Wang, Jianan and
Tan, Wei and
Chen, Yongrui and
Qi, Guilin and
Hao, Wang",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.493",
doi = "10.18653/v1/2023.findings-emnlp.493",
pages = "7373--7383",
abstract = "Table Understanding (TU) is a crucial aspect of information extraction that enables machines to comprehend the semantics behind tabular data. However, existing methods of TU cannot deal with the scarcity of labeled tabular data. In addition, these methods primarily focus on the textual content within the table, disregarding the inherent topological information of the table. This can lead to a misunderstanding of the tabular semantics. In this paper, we propose TabPrompt, a new framework to tackle the above challenges. Prompt-based learning has gained popularity due to its exceptional performance in few-shot learning. Thus, we introduce prompt-based learning to handle few-shot TU. Furthermore, Graph Contrastive Learning (Graph CL) demonstrates remarkable capabilities in capturing topological information, making Graph Neural Networks an ideal method for encoding tables. Hence, we develop a novel Graph CL method tailored to tabular data. This method serves as the pretext task during the pre-training phase, allowing the generation of vector representations that incorporate the table{'}s topological information. The experimental results of outperforming all strong baselines demonstrate the strength of our method in few-shot table understanding tasks.",
}
| Table Understanding (TU) is a crucial aspect of information extraction that enables machines to comprehend the semantics behind tabular data. However, existing methods of TU cannot deal with the scarcity of labeled tabular data. In addition, these methods primarily focus on the textual content within the table, disregarding the inherent topological information of the table. This can lead to a misunderstanding of the tabular semantics. In this paper, we propose TabPrompt, a new framework to tackle the above challenges. Prompt-based learning has gained popularity due to its exceptional performance in few-shot learning. Thus, we introduce prompt-based learning to handle few-shot TU. Furthermore, Graph Contrastive Learning (Graph CL) demonstrates remarkable capabilities in capturing topological information, making Graph Neural Networks an ideal method for encoding tables. Hence, we develop a novel Graph CL method tailored to tabular data. This method serves as the pretext task during the pre-training phase, allowing the generation of vector representations that incorporate the table{'}s topological information. The experimental results of outperforming all strong baselines demonstrate the strength of our method in few-shot table understanding tasks. | [
"Jin, Rihui",
"Wang, Jianan",
"Tan, Wei",
"Chen, Yongrui",
"Qi, Guilin",
"Hao, Wang"
] | TabPrompt: Graph-based Pre-training and Prompting for Few-shot Table Understanding | findings-emnlp.493 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.494.bib | https://aclanthology.org/2023.findings-emnlp.494/ | @inproceedings{kim-etal-2023-towards,
title = "Towards Formality-Aware Neural Machine Translation by Leveraging Context Information",
author = "Kim, Dohee and
Baek, Yujin and
Yang, Soyoung and
Choo, Jaegul",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.494",
doi = "10.18653/v1/2023.findings-emnlp.494",
pages = "7384--7392",
abstract = "Formality is one of the most important linguistic properties to determine the naturalness of translation. Although a target-side context contains formality-related tokens, the sparsity within the context makes it difficult for context-aware neural machine translation (NMT) models to properly discern them. In this paper, we introduce a novel training method to explicitly inform the NMT model by pinpointing key informative tokens using a formality classifier. Given a target context, the formality classifier guides the model to concentrate on the formality-related tokens within the context. Additionally, we modify the standard cross-entropy loss, especially toward the formality-related tokens obtained from the classifier. Experimental results show that our approaches not only improve overall translation quality but also reflect the appropriate formality from the target context.",
}
| Formality is one of the most important linguistic properties to determine the naturalness of translation. Although a target-side context contains formality-related tokens, the sparsity within the context makes it difficult for context-aware neural machine translation (NMT) models to properly discern them. In this paper, we introduce a novel training method to explicitly inform the NMT model by pinpointing key informative tokens using a formality classifier. Given a target context, the formality classifier guides the model to concentrate on the formality-related tokens within the context. Additionally, we modify the standard cross-entropy loss, especially toward the formality-related tokens obtained from the classifier. Experimental results show that our approaches not only improve overall translation quality but also reflect the appropriate formality from the target context. | [
"Kim, Dohee",
"Baek, Yujin",
"Yang, Soyoung",
"Choo, Jaegul"
] | Towards Formality-Aware Neural Machine Translation by Leveraging Context Information | findings-emnlp.494 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.495.bib | https://aclanthology.org/2023.findings-emnlp.495/ | @inproceedings{zhou-etal-2023-improving-seq2seq,
title = "Improving {S}eq2{S}eq Grammatical Error Correction via Decoding Interventions",
author = "Zhou, Houquan and
Liu, Yumeng and
Li, Zhenghua and
Zhang, Min and
Zhang, Bo and
Li, Chen and
Zhang, Ji and
Huang, Fei",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.495",
doi = "10.18653/v1/2023.findings-emnlp.495",
pages = "7393--7405",
abstract = "The sequence-to-sequence (Seq2Seq) approach has recently been widely used in grammatical error correction (GEC) and shows promising performance. However, the Seq2Seq GEC approach still suffers from two issues. First, a Seq2Seq GEC model can only be trained on parallel data, which, in GEC task, is often noisy and limited in quantity. Second, the decoder of a Seq2Seq GEC model lacks an explicit awareness of the correctness of the token being generated. In this paper, we propose a unified decoding intervention framework that employs an external critic to assess the appropriateness of the token to be generated incrementally, and then dynamically influence the choice of the next token. We discover and investigate two types of critics: a pre-trained left-to-right language model critic and an incremental target-side grammatical error detector critic. Through extensive experiments on English and Chinese datasets, our framework consistently outperforms strong baselines and achieves results competitive with state-of-the-art methods.",
}
| The sequence-to-sequence (Seq2Seq) approach has recently been widely used in grammatical error correction (GEC) and shows promising performance. However, the Seq2Seq GEC approach still suffers from two issues. First, a Seq2Seq GEC model can only be trained on parallel data, which, in GEC task, is often noisy and limited in quantity. Second, the decoder of a Seq2Seq GEC model lacks an explicit awareness of the correctness of the token being generated. In this paper, we propose a unified decoding intervention framework that employs an external critic to assess the appropriateness of the token to be generated incrementally, and then dynamically influence the choice of the next token. We discover and investigate two types of critics: a pre-trained left-to-right language model critic and an incremental target-side grammatical error detector critic. Through extensive experiments on English and Chinese datasets, our framework consistently outperforms strong baselines and achieves results competitive with state-of-the-art methods. | [
"Zhou, Houquan",
"Liu, Yumeng",
"Li, Zhenghua",
"Zhang, Min",
"Zhang, Bo",
"Li, Chen",
"Zhang, Ji",
"Huang, Fei"
] | Improving Seq2Seq Grammatical Error Correction via Decoding Interventions | findings-emnlp.495 | 2310.14534 | [
"https://github.com/Jacob-Zhou/gecdi"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.496.bib | https://aclanthology.org/2023.findings-emnlp.496/ | @inproceedings{fan-etal-2023-exploring,
title = "Exploring the Potential of Large Language Models in Generating Code-Tracing Questions for Introductory Programming Courses",
author = "Fan, Aysa and
Zhang, Haoran and
Paquette, Luc and
Zhang, Rui",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.496",
doi = "10.18653/v1/2023.findings-emnlp.496",
pages = "7406--7421",
abstract = "In this paper, we explore the application of large language models (LLMs) for generating code-tracing questions in introductory programming courses. We designed targeted prompts for GPT4, guiding it to generate code-tracing questions based on code snippets and descriptions. We established a set of human evaluation metrics to assess the quality of questions produced by the model compared to those created by human experts. Our analysis provides insights into the capabilities and potential of LLMs in generating diverse code-tracing questions. Additionally, we present a unique dataset of human and LLM-generated tracing questions, serving as a valuable resource for both the education and NLP research communities. This work contributes to the ongoing dialogue on the potential uses of LLMs in educational settings.",
}
| In this paper, we explore the application of large language models (LLMs) for generating code-tracing questions in introductory programming courses. We designed targeted prompts for GPT4, guiding it to generate code-tracing questions based on code snippets and descriptions. We established a set of human evaluation metrics to assess the quality of questions produced by the model compared to those created by human experts. Our analysis provides insights into the capabilities and potential of LLMs in generating diverse code-tracing questions. Additionally, we present a unique dataset of human and LLM-generated tracing questions, serving as a valuable resource for both the education and NLP research communities. This work contributes to the ongoing dialogue on the potential uses of LLMs in educational settings. | [
"Fan, Aysa",
"Zhang, Haoran",
"Paquette, Luc",
"Zhang, Rui"
] | Exploring the Potential of Large Language Models in Generating Code-Tracing Questions for Introductory Programming Courses | findings-emnlp.496 | 2310.15317 | [
"https://github.com/aysafanxm/llm_code_tracing_question_generation"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.497.bib | https://aclanthology.org/2023.findings-emnlp.497/ | @inproceedings{huang-etal-2023-learning-easily,
title = "Learning Easily Updated General Purpose Text Representations with Adaptable Task-Specific Prefix",
author = "Huang, Kuan-Hao and
Tan, Liang and
Hou, Rui and
Wang, Sinong and
Almahairi, Amjad and
Rinott, Ruty",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.497",
doi = "10.18653/v1/2023.findings-emnlp.497",
pages = "7422--7430",
abstract = "Many real-world applications require making multiple predictions from the same text. Fine-tuning a large pre-trained language model for each downstream task causes computational burdens in the inference time due to several times of forward passes. To amortize the computational cost, freezing the language model and building lightweight models for downstream tasks based on fixed text representations are common solutions. Accordingly, how to learn fixed but general text representations that can generalize well to unseen downstream tasks becomes a challenge. Previous works have shown that the generalizability of representations can be improved by fine-tuning the pre-trained language model with some source tasks in a multi-tasking way. In this work, we propose a prefix-based method to learn the fixed text representations with source tasks. We learn a task-specific prefix for each source task independently and combine them to get the final representations. Our experimental results show that prefix-based training performs better than multi-tasking training and can update the text representations at a smaller computational cost than multi-tasking training.",
}
| Many real-world applications require making multiple predictions from the same text. Fine-tuning a large pre-trained language model for each downstream task causes computational burdens in the inference time due to several times of forward passes. To amortize the computational cost, freezing the language model and building lightweight models for downstream tasks based on fixed text representations are common solutions. Accordingly, how to learn fixed but general text representations that can generalize well to unseen downstream tasks becomes a challenge. Previous works have shown that the generalizability of representations can be improved by fine-tuning the pre-trained language model with some source tasks in a multi-tasking way. In this work, we propose a prefix-based method to learn the fixed text representations with source tasks. We learn a task-specific prefix for each source task independently and combine them to get the final representations. Our experimental results show that prefix-based training performs better than multi-tasking training and can update the text representations at a smaller computational cost than multi-tasking training. | [
"Huang, Kuan-Hao",
"Tan, Liang",
"Hou, Rui",
"Wang, Sinong",
"Almahairi, Amjad",
"Rinott, Ruty"
] | Learning Easily Updated General Purpose Text Representations with Adaptable Task-Specific Prefix | findings-emnlp.497 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.498.bib | https://aclanthology.org/2023.findings-emnlp.498/ | @inproceedings{wu-etal-2023-good,
title = "Good Meta-tasks Make A Better Cross-lingual Meta-transfer Learning for Low-resource Languages",
author = "Wu, Linjuan and
Guo, Zongyi and
Cui, Baoliang and
Tang, Haihong and
Lu, Weiming",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.498",
doi = "10.18653/v1/2023.findings-emnlp.498",
pages = "7431--7446",
abstract = "Model-agnostic meta-learning has garnered attention as a promising technique for enhancing few-shot cross-lingual transfer learning in low-resource scenarios. However, little attention was paid to the impact of data selection strategies on this cross-lingual meta-transfer method, particularly the sampling of cross-lingual meta-training data (i.e. meta-tasks) at the syntactic level to reduce language gaps. In this paper, we propose a Meta-Task Collector-based Cross-lingual Meta-Transfer framework (MeTaCo-XMT) to adapt different data selection strategies to construct meta-tasks for meta-transfer learning. Syntactic differences have an effect on transfer performance, so we consider a syntactic similarity sampling strategy and propose a syntactic distance metric model consisting of a syntactic encoder block based on the pre-trained model and a distance metric block using Word Move{'}s Distance (WMD). Additionally, we conduct experiments with three different data selection strategies to instantiate our framework and analyze their performance impact. Experimental results on two multilingual NLP datasets, Wikiann and TydiQA, demonstrate the significant superiority of our approach compared to existing strong baselines.",
}
| Model-agnostic meta-learning has garnered attention as a promising technique for enhancing few-shot cross-lingual transfer learning in low-resource scenarios. However, little attention was paid to the impact of data selection strategies on this cross-lingual meta-transfer method, particularly the sampling of cross-lingual meta-training data (i.e. meta-tasks) at the syntactic level to reduce language gaps. In this paper, we propose a Meta-Task Collector-based Cross-lingual Meta-Transfer framework (MeTaCo-XMT) to adapt different data selection strategies to construct meta-tasks for meta-transfer learning. Syntactic differences have an effect on transfer performance, so we consider a syntactic similarity sampling strategy and propose a syntactic distance metric model consisting of a syntactic encoder block based on the pre-trained model and a distance metric block using Word Move{'}s Distance (WMD). Additionally, we conduct experiments with three different data selection strategies to instantiate our framework and analyze their performance impact. Experimental results on two multilingual NLP datasets, Wikiann and TydiQA, demonstrate the significant superiority of our approach compared to existing strong baselines. | [
"Wu, Linjuan",
"Guo, Zongyi",
"Cui, Baoliang",
"Tang, Haihong",
"Lu, Weiming"
] | Good Meta-tasks Make A Better Cross-lingual Meta-transfer Learning for Low-resource Languages | findings-emnlp.498 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.499.bib | https://aclanthology.org/2023.findings-emnlp.499/ | @inproceedings{lu-etal-2023-reasoning,
title = "Reasoning Makes Good Annotators : An Automatic Task-specific Rules Distilling Framework for Low-resource Relation Extraction",
author = "Lu, Yilin and
Li, Juncheng and
Wang, Xiaoqiang and
Shi, Haochen and
Chen, Tao and
Tang, Siliang",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.499",
doi = "10.18653/v1/2023.findings-emnlp.499",
pages = "7447--7457",
abstract = "Relation extraction is often challenged by insufficient labeled data. Previous methods exploit knowledge from unlabeled data by generating pseudo labels in a self-training pipeline, which suffers a gradual drift problem. Logic rules, a transferable and explainable form of expert knowledge, have achieved promising success by improving the model with weak labels. But manually writing comprehensive rules set is challenging and tedious. To alleviate the human labor of writing high-quality rules, in this work, we propose ARIA, an Automatic task-specific Rules distilling framework. Specifically, we guide the pre-trained language model to reason rules as experts and compose them into robust compound rules for data labeling. Besides, ARIA could continuously enrich the rules set to power the labeling ability by discovering reliable model-labeled data for distinguishable rules generation. Experiments on two public datasets demonstrate the effectiveness of ARIA in a low-resource scenario.",
}
| Relation extraction is often challenged by insufficient labeled data. Previous methods exploit knowledge from unlabeled data by generating pseudo labels in a self-training pipeline, which suffers a gradual drift problem. Logic rules, a transferable and explainable form of expert knowledge, have achieved promising success by improving the model with weak labels. But manually writing comprehensive rules set is challenging and tedious. To alleviate the human labor of writing high-quality rules, in this work, we propose ARIA, an Automatic task-specific Rules distilling framework. Specifically, we guide the pre-trained language model to reason rules as experts and compose them into robust compound rules for data labeling. Besides, ARIA could continuously enrich the rules set to power the labeling ability by discovering reliable model-labeled data for distinguishable rules generation. Experiments on two public datasets demonstrate the effectiveness of ARIA in a low-resource scenario. | [
"Lu, Yilin",
"Li, Juncheng",
"Wang, Xiaoqiang",
"Shi, Haochen",
"Chen, Tao",
"Tang, Siliang"
] | Reasoning Makes Good Annotators : An Automatic Task-specific Rules Distilling Framework for Low-resource Relation Extraction | findings-emnlp.499 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.500.bib | https://aclanthology.org/2023.findings-emnlp.500/ | @inproceedings{lee-etal-2023-co,
title = "Co-training and Co-distillation for Quality Improvement and Compression of Language Models",
author = "Lee, Hayeon and
Hou, Rui and
Kim, Jongpil and
Liang, Davis and
Zhang, Hongbo and
Hwang, Sung and
Min, Alexander",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.500",
doi = "10.18653/v1/2023.findings-emnlp.500",
pages = "7458--7467",
abstract = "Knowledge Distillation (KD) compresses computationally expensive pre-trained language models (PLMs) by transferring their knowledge to smaller models, allowing their use in resource-constrained or real-time settings. However, most smaller models fail to surpass the performance of the original larger model, resulting in sacrificing performance to improve inference speed. To address this issue, we propose Co-Training and Co-Distillation (CTCD), a novel framework that improves performance and inference speed together by co-training two models while mutually distilling knowledge. The CTCD framework successfully achieves this based on two significant findings: 1) Distilling knowledge from the smaller model to the larger model during co-training improves the performance of the larger model. 2) The enhanced performance of the larger model further boosts the performance of the smaller model. The CTCD framework shows promise as it can be combined with existing techniques like architecture design or data augmentation, replacing one-way KD methods, to achieve further performance improvement. Extensive ablation studies demonstrate the effectiveness of CTCD, and the small model distilled by CTCD outperforms the original larger model by a significant margin of 1.66 on the GLUE benchmark.",
}
| Knowledge Distillation (KD) compresses computationally expensive pre-trained language models (PLMs) by transferring their knowledge to smaller models, allowing their use in resource-constrained or real-time settings. However, most smaller models fail to surpass the performance of the original larger model, resulting in sacrificing performance to improve inference speed. To address this issue, we propose Co-Training and Co-Distillation (CTCD), a novel framework that improves performance and inference speed together by co-training two models while mutually distilling knowledge. The CTCD framework successfully achieves this based on two significant findings: 1) Distilling knowledge from the smaller model to the larger model during co-training improves the performance of the larger model. 2) The enhanced performance of the larger model further boosts the performance of the smaller model. The CTCD framework shows promise as it can be combined with existing techniques like architecture design or data augmentation, replacing one-way KD methods, to achieve further performance improvement. Extensive ablation studies demonstrate the effectiveness of CTCD, and the small model distilled by CTCD outperforms the original larger model by a significant margin of 1.66 on the GLUE benchmark. | [
"Lee, Hayeon",
"Hou, Rui",
"Kim, Jongpil",
"Liang, Davis",
"Zhang, Hongbo",
"Hwang, Sung",
"Min, Alex",
"er"
] | Co-training and Co-distillation for Quality Improvement and Compression of Language Models | findings-emnlp.500 | 2311.02849 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.501.bib | https://aclanthology.org/2023.findings-emnlp.501/ | @inproceedings{wang-etal-2023-readprompt,
title = "{R}ead{P}rompt: A Readable Prompting Method for Reliable Knowledge Probing",
author = "Wang, Zezhong and
Ye, Luyao and
Wang, Hongru and
Kwan, Wai-Chung and
Ho, David and
Wong, Kam-Fai",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.501",
doi = "10.18653/v1/2023.findings-emnlp.501",
pages = "7468--7479",
abstract = "Knowledge probing is a task to assess the knowledge encoded within pre-trained language models (PLMs) by having the PLM complete prompts such as {``}Italy is located in {\_}{\_},{''}. The model{'}s prediction precision serves as a lower bound for the amount of knowledge it contains. Subsequent works explore training a series of vectors as prompts to guide PLMs towards more accurate predictions. However, these methods compromise the readability of the prompts. We cannot directly understand these prompts from their literal meaning, making it difficult to verify whether they are correct. Consequently, the credibility of probing results derived from these prompts is diminished. To address the issue, we propose a novel method called ReadPrompt, which aims to identify meaningful sentences to serve as prompts. Experiments show that ReadPrompt achieves state-of-the-art performance on the current knowledge probing benchmark. Moreover, since the prompt is readable, we discovered a misalignment between constructed prompts and knowledge, which is also present in current prompting methods verified by an attack experiment. We claim that the probing outcomes of the current prompting methods are unreliable that overestimate the knowledge contained within PLMs.",
}
| Knowledge probing is a task to assess the knowledge encoded within pre-trained language models (PLMs) by having the PLM complete prompts such as {``}Italy is located in {\_}{\_},{''}. The model{'}s prediction precision serves as a lower bound for the amount of knowledge it contains. Subsequent works explore training a series of vectors as prompts to guide PLMs towards more accurate predictions. However, these methods compromise the readability of the prompts. We cannot directly understand these prompts from their literal meaning, making it difficult to verify whether they are correct. Consequently, the credibility of probing results derived from these prompts is diminished. To address the issue, we propose a novel method called ReadPrompt, which aims to identify meaningful sentences to serve as prompts. Experiments show that ReadPrompt achieves state-of-the-art performance on the current knowledge probing benchmark. Moreover, since the prompt is readable, we discovered a misalignment between constructed prompts and knowledge, which is also present in current prompting methods verified by an attack experiment. We claim that the probing outcomes of the current prompting methods are unreliable that overestimate the knowledge contained within PLMs. | [
"Wang, Zezhong",
"Ye, Luyao",
"Wang, Hongru",
"Kwan, Wai-Chung",
"Ho, David",
"Wong, Kam-Fai"
] | ReadPrompt: A Readable Prompting Method for Reliable Knowledge Probing | findings-emnlp.501 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.502.bib | https://aclanthology.org/2023.findings-emnlp.502/ | @inproceedings{xiao-etal-2023-coherent,
title = "Coherent Entity Disambiguation via Modeling Topic and Categorical Dependency",
author = "Xiao, Zilin and
Shou, Linjun and
Zhang, Xingyao and
Wu, Jie and
Gong, Ming and
Jiang, Daxin",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.502",
doi = "10.18653/v1/2023.findings-emnlp.502",
pages = "7480--7492",
abstract = "Previous entity disambiguation (ED) methods adopt a discriminative paradigm, where prediction is made based on matching scores between mention context and candidate entities using length-limited encoders. However, these methods often struggle to capture explicit discourse-level dependencies, resulting in incoherent predictions at the abstract level (e.g. topic or category). We propose CoherentED, an ED system equipped with novel designs aimed at enhancing the coherence of entity predictions. Our method first introduces an unsupervised variational autoencoder (VAE) to extract latent topic vectors of context sentences. This approach not only allows the encoder to handle longer documents more effectively, conserves valuable input space, but also keeps a topic-level coherence. Additionally, we incorporate an external category memory, enabling the system to retrieve relevant categories for undecided mentions. By employing step-by-step entity decisions, this design facilitates the modeling of entity-entity interactions, thereby maintaining maximum coherence at the category level. We achieve new state-of-the-art results on popular ED benchmarks, with an average improvement of 1.3 F1 points. Our model demonstrates particularly outstanding performance on challenging long-text scenarios.",
}
| Previous entity disambiguation (ED) methods adopt a discriminative paradigm, where prediction is made based on matching scores between mention context and candidate entities using length-limited encoders. However, these methods often struggle to capture explicit discourse-level dependencies, resulting in incoherent predictions at the abstract level (e.g. topic or category). We propose CoherentED, an ED system equipped with novel designs aimed at enhancing the coherence of entity predictions. Our method first introduces an unsupervised variational autoencoder (VAE) to extract latent topic vectors of context sentences. This approach not only allows the encoder to handle longer documents more effectively, conserves valuable input space, but also keeps a topic-level coherence. Additionally, we incorporate an external category memory, enabling the system to retrieve relevant categories for undecided mentions. By employing step-by-step entity decisions, this design facilitates the modeling of entity-entity interactions, thereby maintaining maximum coherence at the category level. We achieve new state-of-the-art results on popular ED benchmarks, with an average improvement of 1.3 F1 points. Our model demonstrates particularly outstanding performance on challenging long-text scenarios. | [
"Xiao, Zilin",
"Shou, Linjun",
"Zhang, Xingyao",
"Wu, Jie",
"Gong, Ming",
"Jiang, Daxin"
] | Coherent Entity Disambiguation via Modeling Topic and Categorical Dependency | findings-emnlp.502 | 2311.03253 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.503.bib | https://aclanthology.org/2023.findings-emnlp.503/ | @inproceedings{ye-etal-2023-predictable,
title = "How Predictable Are Large Language Model Capabilities? A Case Study on {BIG}-bench",
author = "Ye, Qinyuan and
Fu, Harvey and
Ren, Xiang and
Jia, Robin",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.503",
doi = "10.18653/v1/2023.findings-emnlp.503",
pages = "7493--7517",
abstract = "We investigate the predictability of large language model (LLM) capabilities: given records of past experiments using different model families, numbers of parameters, tasks, and numbers of in-context examples, can we accurately predict LLM performance on new experiment configurations? Answering this question has practical implications for LLM users (e.g., deciding which models to try), developers (e.g., prioritizing evaluation on representative tasks), and the research community (e.g., identifying hard-to-predict capabilities that warrant further investigation). We study the performance prediction problem on experiment records from BIG-bench. On a random train-test split, an MLP-based predictor achieves an $R^2$ score greater than 95{\%}, indicating the presence of learnable patterns within the experiment records. We then formulate the problem of searching for {``}small-bench,{''} an informative subset of BIG-bench tasks from which the performance on the full set can be maximally recovered. We find a subset as informative as BIG-bench Hard for evaluating new model families, while being $3\times$ smaller. Additionally, we find competitive subsets by clustering task representations learned by our MLP-based predictor and selecting tasks close to cluster centroids, highlighting the importance of task diversity in constructing {``}small-bench.{''}",
}
| We investigate the predictability of large language model (LLM) capabilities: given records of past experiments using different model families, numbers of parameters, tasks, and numbers of in-context examples, can we accurately predict LLM performance on new experiment configurations? Answering this question has practical implications for LLM users (e.g., deciding which models to try), developers (e.g., prioritizing evaluation on representative tasks), and the research community (e.g., identifying hard-to-predict capabilities that warrant further investigation). We study the performance prediction problem on experiment records from BIG-bench. On a random train-test split, an MLP-based predictor achieves an $R^2$ score greater than 95{\%}, indicating the presence of learnable patterns within the experiment records. We then formulate the problem of searching for {``}small-bench,{''} an informative subset of BIG-bench tasks from which the performance on the full set can be maximally recovered. We find a subset as informative as BIG-bench Hard for evaluating new model families, while being $3\times$ smaller. Additionally, we find competitive subsets by clustering task representations learned by our MLP-based predictor and selecting tasks close to cluster centroids, highlighting the importance of task diversity in constructing {``}small-bench.{''} | [
"Ye, Qinyuan",
"Fu, Harvey",
"Ren, Xiang",
"Jia, Robin"
] | How Predictable Are Large Language Model Capabilities? A Case Study on BIG-bench | findings-emnlp.503 | 2305.14947 | [
"https://github.com/ink-usc/predicting-big-bench"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.504.bib | https://aclanthology.org/2023.findings-emnlp.504/ | @inproceedings{shu-etal-2023-posqa,
title = "{POSQA}: Probe the World Models of {LLM}s with Size Comparisons",
author = "Shu, Chang and
Han, Jiuzhou and
Liu, Fangyu and
Shareghi, Ehsan and
Collier, Nigel",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.504",
doi = "10.18653/v1/2023.findings-emnlp.504",
pages = "7518--7531",
abstract = "Embodied language comprehension emphasizes that language understanding is not solely a matter of mental processing in the brain but also involves interactions with the physical and social environment. With the explosive growth of Large Language Models (LLMs) and their already ubiquitous presence in our daily lives, it is becoming increasingly necessary to verify their real-world understanding. Inspired by cognitive theories, we propose POSQA: a Physical Object Size Question Answering dataset with simple size comparison questions to examine the extremity and analyze the potential mechanisms of the embodied comprehension of the latest LLMs. We show that even the largest LLMs today perform poorly under the zero-shot setting. We then push their limits with advanced prompting techniques and external knowledge augmentation. Furthermore, we investigate whether their real-world comprehension primarily derives from contextual information or internal weights and analyse the impact of prompt formats and report bias of different objects. Our results show that real-world understanding that LLMs shaped from textual data can be vulnerable to deception and confusion by the surface form of prompts, which makes it less aligned with human behaviours.",
}
| Embodied language comprehension emphasizes that language understanding is not solely a matter of mental processing in the brain but also involves interactions with the physical and social environment. With the explosive growth of Large Language Models (LLMs) and their already ubiquitous presence in our daily lives, it is becoming increasingly necessary to verify their real-world understanding. Inspired by cognitive theories, we propose POSQA: a Physical Object Size Question Answering dataset with simple size comparison questions to examine the extremity and analyze the potential mechanisms of the embodied comprehension of the latest LLMs. We show that even the largest LLMs today perform poorly under the zero-shot setting. We then push their limits with advanced prompting techniques and external knowledge augmentation. Furthermore, we investigate whether their real-world comprehension primarily derives from contextual information or internal weights and analyse the impact of prompt formats and report bias of different objects. Our results show that real-world understanding that LLMs shaped from textual data can be vulnerable to deception and confusion by the surface form of prompts, which makes it less aligned with human behaviours. | [
"Shu, Chang",
"Han, Jiuzhou",
"Liu, Fangyu",
"Shareghi, Ehsan",
"Collier, Nigel"
] | POSQA: Probe the World Models of LLMs with Size Comparisons | findings-emnlp.504 | 2310.13394 | [
"https://github.com/cambridgeltl/posqa"
] | https://huggingface.co/papers/2310.13394 | 0 | 1 | 0 | 5 | [] | [
"cambridgeltl/posqa"
] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.505.bib | https://aclanthology.org/2023.findings-emnlp.505/ | @inproceedings{miah-etal-2023-hierarchical,
title = "Hierarchical Fusion for Online Multimodal Dialog Act Classification",
author = "Miah, Md Messal Monem and
Pyarelal, Adarsh and
Huang, Ruihong",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.505",
doi = "10.18653/v1/2023.findings-emnlp.505",
pages = "7532--7545",
abstract = "We propose a framework for online multimodal dialog act (DA) classification based on raw audio and ASR-generated transcriptions of current and past utterances. Existing multimodal DA classification approaches are limited by ineffective audio modeling and late-stage fusion. We showcase significant improvements in multimodal DA classification by integrating modalities at a more granular level and incorporating recent advancements in large language and audio models for audio feature extraction. We further investigate the effectiveness of self-attention and cross-attention mechanisms in modeling utterances and dialogs for DA classification. We achieve a substantial increase of 3 percentage points in the F1 score relative to current state-of-the-art models on two prominent DA classification datasets, MRDA and EMOTyDA.",
}
| We propose a framework for online multimodal dialog act (DA) classification based on raw audio and ASR-generated transcriptions of current and past utterances. Existing multimodal DA classification approaches are limited by ineffective audio modeling and late-stage fusion. We showcase significant improvements in multimodal DA classification by integrating modalities at a more granular level and incorporating recent advancements in large language and audio models for audio feature extraction. We further investigate the effectiveness of self-attention and cross-attention mechanisms in modeling utterances and dialogs for DA classification. We achieve a substantial increase of 3 percentage points in the F1 score relative to current state-of-the-art models on two prominent DA classification datasets, MRDA and EMOTyDA. | [
"Miah, Md Messal Monem",
"Pyarelal, Adarsh",
"Huang, Ruihong"
] | Hierarchical Fusion for Online Multimodal Dialog Act Classification | findings-emnlp.505 | [
"https://github.com/Dipto084/Hierarchical-Fusion-for-Online-Multimodal-Dialog-Act-Classification"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.506.bib | https://aclanthology.org/2023.findings-emnlp.506/ | @inproceedings{hallinan-etal-2023-steer,
title = "{STEER}: Unified Style Transfer with Expert Reinforcement",
author = "Hallinan, Skyler and
Brahman, Faeze and
Lu, Ximing and
Jung, Jaehun and
Welleck, Sean and
Choi, Yejin",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.506",
doi = "10.18653/v1/2023.findings-emnlp.506",
pages = "7546--7562",
abstract = "While text style transfer has many applications across natural language processing, the core premise of transferring from a single source style is unrealistic in a real-world setting. In this work, we focus on arbitrary style transfer: rewriting a text from an arbitrary, unknown style to a target style. We propose STEER: Unified Style Transfer with Expert Reinforcement, a unified frame-work developed to overcome the challenge of limited parallel data for style transfer. STEER involves automatically generating a corpus of style-transfer pairs using a product of experts during decoding. The generated offline data is then used to pre-train an initial policy before switching to online, off-policy reinforcement learning for further improvements via fine-grained reward signals. STEER is unified and can transfer to multiple target styles from an arbitrary, unknown source style, making it particularly flexible and efficient. Experimental results on a challenging dataset with text from a diverse set of styles demonstrate state-of-the-art results compared to competitive baselines. Remarkably, STEER outperforms the 175B parameter instruction-tuned GPT-3 on overall style transfer quality, despite being 226 times smaller in size. We also show STEER is robust, maintaining its style transfer capabilities on out-of-domain data, and surpassing nearly all baselines across various styles. The success of our method highlights the potential of RL algorithms when augmented with controllable decoding to overcome the challenge of limited data supervision.",
}
| While text style transfer has many applications across natural language processing, the core premise of transferring from a single source style is unrealistic in a real-world setting. In this work, we focus on arbitrary style transfer: rewriting a text from an arbitrary, unknown style to a target style. We propose STEER: Unified Style Transfer with Expert Reinforcement, a unified frame-work developed to overcome the challenge of limited parallel data for style transfer. STEER involves automatically generating a corpus of style-transfer pairs using a product of experts during decoding. The generated offline data is then used to pre-train an initial policy before switching to online, off-policy reinforcement learning for further improvements via fine-grained reward signals. STEER is unified and can transfer to multiple target styles from an arbitrary, unknown source style, making it particularly flexible and efficient. Experimental results on a challenging dataset with text from a diverse set of styles demonstrate state-of-the-art results compared to competitive baselines. Remarkably, STEER outperforms the 175B parameter instruction-tuned GPT-3 on overall style transfer quality, despite being 226 times smaller in size. We also show STEER is robust, maintaining its style transfer capabilities on out-of-domain data, and surpassing nearly all baselines across various styles. The success of our method highlights the potential of RL algorithms when augmented with controllable decoding to overcome the challenge of limited data supervision. | [
"Hallinan, Skyler",
"Brahman, Faeze",
"Lu, Ximing",
"Jung, Jaehun",
"Welleck, Sean",
"Choi, Yejin"
] | STEER: Unified Style Transfer with Expert Reinforcement | findings-emnlp.506 | 2311.07167 | [
"https://github.com/shallinan1/steerstyletransfer"
] | https://huggingface.co/papers/2311.07167 | 1 | 0 | 0 | 6 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.findings-emnlp.507.bib | https://aclanthology.org/2023.findings-emnlp.507/ | @inproceedings{luo-etal-2023-enhancing,
title = "Enhancing Argument Structure Extraction with Efficient Leverage of Contextual Information",
author = "Luo, Yun and
Yang, Zhen and
Meng, Fandong and
Li, Yingjie and
Zhou, Jie and
Zhang, Yue",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.507",
doi = "10.18653/v1/2023.findings-emnlp.507",
pages = "7563--7571",
abstract = "Argument structure extraction (ASE) aims to identify the discourse structure of arguments within documents. Previous research has demonstrated that contextual information is crucial for developing an effective ASE model. However, we observe that merely concatenating sentences in a contextual window does not fully utilize contextual information and can sometimes lead to excessive attention on less informative sentences. To tackle this challenge, we propose an Efficient Context-aware ASE model (ECASE) that fully exploits contextual information by enhancing modeling capacity and augmenting training data. Specifically, we introduce a sequence-attention module and distance-weighted similarity loss to aggregate contextual information and argumentative information. Additionally, we augment the training data by randomly masking discourse markers and sentences, which reduces the model{'}s reliance on specific words or less informative sentences. Our experiments on five datasets from various domains demonstrate that our model achieves state-of-the-art performance. Furthermore, ablation studies confirm the effectiveness of each module in our model.",
}
| Argument structure extraction (ASE) aims to identify the discourse structure of arguments within documents. Previous research has demonstrated that contextual information is crucial for developing an effective ASE model. However, we observe that merely concatenating sentences in a contextual window does not fully utilize contextual information and can sometimes lead to excessive attention on less informative sentences. To tackle this challenge, we propose an Efficient Context-aware ASE model (ECASE) that fully exploits contextual information by enhancing modeling capacity and augmenting training data. Specifically, we introduce a sequence-attention module and distance-weighted similarity loss to aggregate contextual information and argumentative information. Additionally, we augment the training data by randomly masking discourse markers and sentences, which reduces the model{'}s reliance on specific words or less informative sentences. Our experiments on five datasets from various domains demonstrate that our model achieves state-of-the-art performance. Furthermore, ablation studies confirm the effectiveness of each module in our model. | [
"Luo, Yun",
"Yang, Zhen",
"Meng, F",
"ong",
"Li, Yingjie",
"Zhou, Jie",
"Zhang, Yue"
] | Enhancing Argument Structure Extraction with Efficient Leverage of Contextual Information | findings-emnlp.507 | 2310.05073 | [
"https://github.com/luoxiaoheics/ecase"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.508.bib | https://aclanthology.org/2023.findings-emnlp.508/ | @inproceedings{xiong-etal-2023-examining,
title = "Examining Inter-Consistency of Large Language Models Collaboration: An In-depth Analysis via Debate",
author = "Xiong, Kai and
Ding, Xiao and
Cao, Yixin and
Liu, Ting and
Qin, Bing",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.508",
doi = "10.18653/v1/2023.findings-emnlp.508",
pages = "7572--7590",
abstract = "Large Language Models (LLMs) have shown impressive capabilities in various applications, but they still face various inconsistency issues. Existing works primarily focus on the inconsistency issues within a single LLM, while we complementarily explore the inter-consistency among multiple LLMs for collaboration. To examine whether LLMs can collaborate effectively to achieve a consensus for a shared goal, we focus on commonsense reasoning, and introduce a formal debate framework (FORD) to conduct a three-stage debate among LLMs with real-world scenarios alignment: fair debate, mismatched debate, and roundtable debate. Through extensive experiments on various datasets, LLMs can effectively collaborate to reach a consensus despite noticeable inter-inconsistencies, but imbalances in their abilities can lead to domination by superior LLMs. Leveraging a more advanced LLM like GPT-4 as an authoritative judge can boost collaboration performance. Our work contributes to understanding the inter-consistency among LLMs and lays the foundation for developing future collaboration methods. Codes and data are available at https://github.com/Waste-Wood/FORD.",
}
| Large Language Models (LLMs) have shown impressive capabilities in various applications, but they still face various inconsistency issues. Existing works primarily focus on the inconsistency issues within a single LLM, while we complementarily explore the inter-consistency among multiple LLMs for collaboration. To examine whether LLMs can collaborate effectively to achieve a consensus for a shared goal, we focus on commonsense reasoning, and introduce a formal debate framework (FORD) to conduct a three-stage debate among LLMs with real-world scenarios alignment: fair debate, mismatched debate, and roundtable debate. Through extensive experiments on various datasets, LLMs can effectively collaborate to reach a consensus despite noticeable inter-inconsistencies, but imbalances in their abilities can lead to domination by superior LLMs. Leveraging a more advanced LLM like GPT-4 as an authoritative judge can boost collaboration performance. Our work contributes to understanding the inter-consistency among LLMs and lays the foundation for developing future collaboration methods. Codes and data are available at https://github.com/Waste-Wood/FORD. | [
"Xiong, Kai",
"Ding, Xiao",
"Cao, Yixin",
"Liu, Ting",
"Qin, Bing"
] | Examining Inter-Consistency of Large Language Models Collaboration: An In-depth Analysis via Debate | findings-emnlp.508 | 2305.11595 | [
"https://github.com/waste-wood/ford"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.509.bib | https://aclanthology.org/2023.findings-emnlp.509/ | @inproceedings{huang-yang-2023-culturally,
title = "Culturally Aware Natural Language Inference",
author = "Huang, Jing and
Yang, Diyi",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.509",
doi = "10.18653/v1/2023.findings-emnlp.509",
pages = "7591--7609",
abstract = "Humans produce and consume language in a particular cultural context, which includes knowledge about specific norms and practices. A listener{'}s awareness of the cultural context is critical for interpreting the speaker{'}s meaning. A simple expression like *I didn{'}t leave a tip* implies a strong sense of dissatisfaction when tipping is assumed to be the norm. As NLP systems reach users from different cultures, achieving culturally aware language understanding becomes increasingly important. However, current research has focused on building cultural knowledge bases without studying how such knowledge leads to contextualized interpretations of texts. In this work, we operationalize cultural variations in language understanding through a natural language inference (NLI) task that surfaces cultural variations as label disagreement between annotators from different cultural groups. We introduce the first Culturally Aware Natural Language Inference (CALI) dataset with 2.7K premise-hypothesis pairs annotated by two cultural groups located in the U.S. and India. With CALI, we categorize how cultural norms affect language understanding and present an evaluation framework to assess at which levels large language models are culturally aware. Our dataset is available at https://github.com/SALT-NLP/CulturallyAwareNLI.",
}
| Humans produce and consume language in a particular cultural context, which includes knowledge about specific norms and practices. A listener{'}s awareness of the cultural context is critical for interpreting the speaker{'}s meaning. A simple expression like *I didn{'}t leave a tip* implies a strong sense of dissatisfaction when tipping is assumed to be the norm. As NLP systems reach users from different cultures, achieving culturally aware language understanding becomes increasingly important. However, current research has focused on building cultural knowledge bases without studying how such knowledge leads to contextualized interpretations of texts. In this work, we operationalize cultural variations in language understanding through a natural language inference (NLI) task that surfaces cultural variations as label disagreement between annotators from different cultural groups. We introduce the first Culturally Aware Natural Language Inference (CALI) dataset with 2.7K premise-hypothesis pairs annotated by two cultural groups located in the U.S. and India. With CALI, we categorize how cultural norms affect language understanding and present an evaluation framework to assess at which levels large language models are culturally aware. Our dataset is available at https://github.com/SALT-NLP/CulturallyAwareNLI. | [
"Huang, Jing",
"Yang, Diyi"
] | Culturally Aware Natural Language Inference | findings-emnlp.509 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.510.bib | https://aclanthology.org/2023.findings-emnlp.510/ | @inproceedings{towle-zhou-2023-end,
title = "End-to-End Autoregressive Retrieval via Bootstrapping for Smart Reply Systems",
author = "Towle, Benjamin and
Zhou, Ke",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.510",
doi = "10.18653/v1/2023.findings-emnlp.510",
pages = "7610--7622",
abstract = "Reply suggestion systems represent a staple component of many instant messaging and email systems. However, the requirement to produce sets of replies, rather than individual replies, makes the task poorly suited for out-of-the-box retrieval architectures, which only consider individual message-reply similarity. As a result, these system often rely on additional post-processing modules to diversify the outputs. However, these approaches are ultimately bottlenecked by the performance of the initial retriever, which in practice struggles to present a sufficiently diverse range of options to the downstream diversification module, leading to the suggestions being less relevant to the user. In this paper, we consider a novel approach that radically simplifies this pipeline through an autoregressive text-to-text retrieval model, that learns the smart reply task end-to-end from a dataset of (message, reply set) pairs obtained via bootstrapping. Empirical results show this method consistently outperforms a range of state-of-the-art baselines across three datasets, corresponding to a 5.1{\%}-17.9{\%} improvement in relevance, and a 0.5{\%}-63.1{\%} improvement in diversity compared to the best baseline approach. We make our code publicly available.",
}
| Reply suggestion systems represent a staple component of many instant messaging and email systems. However, the requirement to produce sets of replies, rather than individual replies, makes the task poorly suited for out-of-the-box retrieval architectures, which only consider individual message-reply similarity. As a result, these system often rely on additional post-processing modules to diversify the outputs. However, these approaches are ultimately bottlenecked by the performance of the initial retriever, which in practice struggles to present a sufficiently diverse range of options to the downstream diversification module, leading to the suggestions being less relevant to the user. In this paper, we consider a novel approach that radically simplifies this pipeline through an autoregressive text-to-text retrieval model, that learns the smart reply task end-to-end from a dataset of (message, reply set) pairs obtained via bootstrapping. Empirical results show this method consistently outperforms a range of state-of-the-art baselines across three datasets, corresponding to a 5.1{\%}-17.9{\%} improvement in relevance, and a 0.5{\%}-63.1{\%} improvement in diversity compared to the best baseline approach. We make our code publicly available. | [
"Towle, Benjamin",
"Zhou, Ke"
] | End-to-End Autoregressive Retrieval via Bootstrapping for Smart Reply Systems | findings-emnlp.510 | 2310.18956 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.511.bib | https://aclanthology.org/2023.findings-emnlp.511/ | @inproceedings{li-etal-2023-evaluating-dependencies,
title = "Evaluating Dependencies in Fact Editing for Language Models: Specificity and Implication Awareness",
author = "Li, Zichao and
Arous, Ines and
Reddy, Siva and
Cheung, Jackie",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.511",
doi = "10.18653/v1/2023.findings-emnlp.511",
pages = "7623--7636",
abstract = "The potential of using a large language model (LLM) as a knowledge base (KB) has sparked significant interest. To maintain the knowledge acquired by LLMs, we need to ensure that the editing of learned facts respects internal logical constraints, which are known as dependency of knowledge. Existing work on editing LLMs has partially addressed the issue of dependency, when the editing of a fact should apply to its lexical variations without disrupting irrelevant ones. However, they neglect the dependency between a fact and its logical implications. We propose an evaluation protocol with an accompanying question-answering dataset, StandUp, that provides a comprehensive assessment of the editing process considering the above notions of dependency. Our protocol involves setting up a controlled environment in which we edit facts and monitor their impact on LLMs, along with their implications based on If-Then rules. Extensive experiments on StandUp show that existing knowledge editing methods are sensitive to the surface form of knowledge, and that they have limited performance in inferring the implications of edited facts.",
}
| The potential of using a large language model (LLM) as a knowledge base (KB) has sparked significant interest. To maintain the knowledge acquired by LLMs, we need to ensure that the editing of learned facts respects internal logical constraints, which are known as dependency of knowledge. Existing work on editing LLMs has partially addressed the issue of dependency, when the editing of a fact should apply to its lexical variations without disrupting irrelevant ones. However, they neglect the dependency between a fact and its logical implications. We propose an evaluation protocol with an accompanying question-answering dataset, StandUp, that provides a comprehensive assessment of the editing process considering the above notions of dependency. Our protocol involves setting up a controlled environment in which we edit facts and monitor their impact on LLMs, along with their implications based on If-Then rules. Extensive experiments on StandUp show that existing knowledge editing methods are sensitive to the surface form of knowledge, and that they have limited performance in inferring the implications of edited facts. | [
"Li, Zichao",
"Arous, Ines",
"Reddy, Siva",
"Cheung, Jackie"
] | Evaluating Dependencies in Fact Editing for Language Models: Specificity and Implication Awareness | findings-emnlp.511 | 2312.01858 | [
"https://github.com/mcgill-nlp/logicalknowedit"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.512.bib | https://aclanthology.org/2023.findings-emnlp.512/ | @inproceedings{elangovan-etal-2023-effects,
title = "Effects of Human Adversarial and Affable Samples on {BERT} Generalization",
author = "Elangovan, Aparna and
He, Estrid and
Li, Yuan and
Verspoor, Karin",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.512",
doi = "10.18653/v1/2023.findings-emnlp.512",
pages = "7637--7649",
abstract = "BERT-based models have had strong performance on leaderboards, yet have been demonstrably worse in real-world settings requiring generalization. Limited quantities of training data is considered a key impediment to achieving generalizability in machine learning. In this paper, we examine the impact of training data quality, not quantity, on a model{'}s generalizability. We consider two characteristics of training data: the portion of human-adversarial (h-adversarial), i.e. sample pairs with seemingly minor differences but different ground-truth labels, and human-affable (h-affable) training samples, i.e. sample pairs with minor differences but the same ground-truth label. We find that for a fixed size of training samples, as a rule of thumb, having 10-30{\%} h-adversarial instances improves the precision, and therefore F1, by up to 20 points in the tasks of text classification and relation extraction. Increasing h-adversarials beyond this range can result in performance plateaus or even degradation. In contrast, h-affables may not contribute to a model{'}s generalizability and may even degrade generalization performance.",
}
| BERT-based models have had strong performance on leaderboards, yet have been demonstrably worse in real-world settings requiring generalization. Limited quantities of training data is considered a key impediment to achieving generalizability in machine learning. In this paper, we examine the impact of training data quality, not quantity, on a model{'}s generalizability. We consider two characteristics of training data: the portion of human-adversarial (h-adversarial), i.e. sample pairs with seemingly minor differences but different ground-truth labels, and human-affable (h-affable) training samples, i.e. sample pairs with minor differences but the same ground-truth label. We find that for a fixed size of training samples, as a rule of thumb, having 10-30{\%} h-adversarial instances improves the precision, and therefore F1, by up to 20 points in the tasks of text classification and relation extraction. Increasing h-adversarials beyond this range can result in performance plateaus or even degradation. In contrast, h-affables may not contribute to a model{'}s generalizability and may even degrade generalization performance. | [
"Elangovan, Aparna",
"He, Estrid",
"Li, Yuan",
"Verspoor, Karin"
] | Effects of Human Adversarial and Affable Samples on BERT Generalization | findings-emnlp.512 | 2310.08008 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.513.bib | https://aclanthology.org/2023.findings-emnlp.513/ | @inproceedings{fu-etal-2023-logic,
title = "Logic Unveils Truth, While Disguise Obscures It: Transition Logic Augmented Response Selection for Multi-Turn Dialogue",
author = "Fu, Tingchen and
Zhao, Xueliang and
Liu, Lemao and
Yan, Rui",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.513",
doi = "10.18653/v1/2023.findings-emnlp.513",
pages = "7650--7661",
abstract = "Multi-turn response selection aims to retrieve a response for a dialogue context from a candidate pool and negative sampling is the key to its retrieval performance. However, previous methods of negative samples tend to yield false negatives due to the one-to-many property in open-domain dialogue, which is detrimental to the optimization process. To deal with the problem, we propose a sequential variational ladder auto-encoder to capture the diverse one-to-many transition pattern of multiple characteristics in open-domain dialogue. The learned transition logic thus assists in identifying potential positives in disguise. Meanwhile, we propose a TRIGGER framework to adjust negative sampling in the training process such that the scope of false negatives dynamically updates according to the model capacity. Extensive experiments on two benchmarks verify the effectiveness of our approach.",
}
| Multi-turn response selection aims to retrieve a response for a dialogue context from a candidate pool and negative sampling is the key to its retrieval performance. However, previous methods of negative samples tend to yield false negatives due to the one-to-many property in open-domain dialogue, which is detrimental to the optimization process. To deal with the problem, we propose a sequential variational ladder auto-encoder to capture the diverse one-to-many transition pattern of multiple characteristics in open-domain dialogue. The learned transition logic thus assists in identifying potential positives in disguise. Meanwhile, we propose a TRIGGER framework to adjust negative sampling in the training process such that the scope of false negatives dynamically updates according to the model capacity. Extensive experiments on two benchmarks verify the effectiveness of our approach. | [
"Fu, Tingchen",
"Zhao, Xueliang",
"Liu, Lemao",
"Yan, Rui"
] | Logic Unveils Truth, While Disguise Obscures It: Transition Logic Augmented Response Selection for Multi-Turn Dialogue | findings-emnlp.513 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.514.bib | https://aclanthology.org/2023.findings-emnlp.514/ | @inproceedings{webson-etal-2023-language,
title = "Are Language Models Worse than Humans at Following Prompts? It{'}s Complicated",
author = "Webson, Albert and
Loo, Alyssa and
Yu, Qinan and
Pavlick, Ellie",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.514",
doi = "10.18653/v1/2023.findings-emnlp.514",
pages = "7662--7686",
abstract = "Prompts have been the center of progress in advancing language models{'} zero-shot and few-shot performance. However, recent work finds that models can perform surprisingly well when given intentionally irrelevant or misleading prompts. Such results may be interpreted as evidence that model behavior is not {``}human like{'}. In this study, we challenge a central assumption in such work: that humans would perform badly when given pathological instructions. We find that humans are able to reliably ignore irrelevant instructions and thus, like models, perform well on the underlying task despite an apparent lack of signal regarding the task they are being asked to do. However, when given deliberately misleading instructions, humans follow the instructions faithfully, whereas models do not. Thus, our conclusion is mixed with respect to prior work. We argue against the earlier claim that high performance with irrelevant prompts constitutes evidence against models{'} instruction understanding, but we reinforce the claim that models{'} failure to follow misleading instructions raises concerns. More broadly, we caution that future research should not idealize human behaviors as a monolith and should not train or evaluate models to mimic assumptions about these behaviors without first validating humans{'} behaviors empirically.",
}
| Prompts have been the center of progress in advancing language models{'} zero-shot and few-shot performance. However, recent work finds that models can perform surprisingly well when given intentionally irrelevant or misleading prompts. Such results may be interpreted as evidence that model behavior is not {``}human like{'}. In this study, we challenge a central assumption in such work: that humans would perform badly when given pathological instructions. We find that humans are able to reliably ignore irrelevant instructions and thus, like models, perform well on the underlying task despite an apparent lack of signal regarding the task they are being asked to do. However, when given deliberately misleading instructions, humans follow the instructions faithfully, whereas models do not. Thus, our conclusion is mixed with respect to prior work. We argue against the earlier claim that high performance with irrelevant prompts constitutes evidence against models{'} instruction understanding, but we reinforce the claim that models{'} failure to follow misleading instructions raises concerns. More broadly, we caution that future research should not idealize human behaviors as a monolith and should not train or evaluate models to mimic assumptions about these behaviors without first validating humans{'} behaviors empirically. | [
"Webson, Albert",
"Loo, Alyssa",
"Yu, Qinan",
"Pavlick, Ellie"
] | Are Language Models Worse than Humans at Following Prompts? It's Complicated | findings-emnlp.514 | 2301.07085 | [
"https://github.com/awebson/prompt_semantics"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.515.bib | https://aclanthology.org/2023.findings-emnlp.515/ | @inproceedings{song-etal-2023-sequence,
title = "A Sequence-to-Structure Approach to Document-level Targeted Sentiment Analysis",
author = "Song, Nan and
Cai, Hongjie and
Xia, Rui and
Yu, Jianfei and
Wu, Zhen and
Dai, Xinyu",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.515",
doi = "10.18653/v1/2023.findings-emnlp.515",
pages = "7687--7698",
abstract = "Most previous studies on aspect-based sentiment analysis (ABSA) were carried out at the sentence level, while the research of document-level ABSA has not received enough attention. In this work, we focus on the document-level targeted sentiment analysis task, which aims to extract the opinion targets consisting of multi-level entities from a review document and predict their sentiments. We propose a Sequence-to-Structure (Seq2Struct) approach to address the task, which is able to explicitly model the hierarchical structure among multiple opinion targets in a document, and capture the long-distance dependencies among affiliated entities across sentences. In addition to the existing Seq2Seq approach, we further construct four strong baselines with different pretrained models. Experimental results on six domains show that our Seq2Struct approach outperforms all the baselines significantly. Aside from the performance advantage in outputting the multi-level target-sentiment pairs, our approach has another significant advantage - it can explicitly display the hierarchical structure of the opinion targets within a document. Our source code is publicly released at https://github.com/NUSTM/Doc-TSA-Seq2Struct.",
}
| Most previous studies on aspect-based sentiment analysis (ABSA) were carried out at the sentence level, while the research of document-level ABSA has not received enough attention. In this work, we focus on the document-level targeted sentiment analysis task, which aims to extract the opinion targets consisting of multi-level entities from a review document and predict their sentiments. We propose a Sequence-to-Structure (Seq2Struct) approach to address the task, which is able to explicitly model the hierarchical structure among multiple opinion targets in a document, and capture the long-distance dependencies among affiliated entities across sentences. In addition to the existing Seq2Seq approach, we further construct four strong baselines with different pretrained models. Experimental results on six domains show that our Seq2Struct approach outperforms all the baselines significantly. Aside from the performance advantage in outputting the multi-level target-sentiment pairs, our approach has another significant advantage - it can explicitly display the hierarchical structure of the opinion targets within a document. Our source code is publicly released at https://github.com/NUSTM/Doc-TSA-Seq2Struct. | [
"Song, Nan",
"Cai, Hongjie",
"Xia, Rui",
"Yu, Jianfei",
"Wu, Zhen",
"Dai, Xinyu"
] | A Sequence-to-Structure Approach to Document-level Targeted Sentiment Analysis | findings-emnlp.515 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.516.bib | https://aclanthology.org/2023.findings-emnlp.516/ | @inproceedings{zhang-yang-2023-generating,
title = "Generating Extractive Answers: Gated Recurrent Memory Reader for Conversational Question Answering",
author = "Zhang, Xuanyu and
Yang, Qing",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.516",
doi = "10.18653/v1/2023.findings-emnlp.516",
pages = "7699--7704",
abstract = "Conversational question answering (CQA) is a more complicated task than traditional single-turn machine reading comprehension (MRC). Different from large language models (LLMs) like ChatGPT, the models of CQA need to extract answers from given contents to answer follow-up questions according to conversation history. In this paper, we propose a novel architecture, i.e., Gated Recurrent Memory Reader (GRMR), which integrates traditional extractive MRC models into a generalized sequence-to-sequence framework. After the passage is encoded, the decoder will generate the extractive answers turn by turn. Different from previous models that concatenate the previous questions and answers as context superficially and redundantly, our model can use less storage space and consider historical memory deeply and selectively. Experiments on the Conversational Question Answering (CoQA) dataset show that our model achieves comparable results to most models with the least space occupancy.",
}
| Conversational question answering (CQA) is a more complicated task than traditional single-turn machine reading comprehension (MRC). Different from large language models (LLMs) like ChatGPT, the models of CQA need to extract answers from given contents to answer follow-up questions according to conversation history. In this paper, we propose a novel architecture, i.e., Gated Recurrent Memory Reader (GRMR), which integrates traditional extractive MRC models into a generalized sequence-to-sequence framework. After the passage is encoded, the decoder will generate the extractive answers turn by turn. Different from previous models that concatenate the previous questions and answers as context superficially and redundantly, our model can use less storage space and consider historical memory deeply and selectively. Experiments on the Conversational Question Answering (CoQA) dataset show that our model achieves comparable results to most models with the least space occupancy. | [
"Zhang, Xuanyu",
"Yang, Qing"
] | Generating Extractive Answers: Gated Recurrent Memory Reader for Conversational Question Answering | findings-emnlp.516 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
||
https://aclanthology.org/2023.findings-emnlp.517.bib | https://aclanthology.org/2023.findings-emnlp.517/ | @inproceedings{yan-etal-2023-text2tree,
title = "{T}ext2{T}ree: Aligning Text Representation to the Label Tree Hierarchy for Imbalanced Medical Classification",
author = "Yan, Jiahuan and
Gao, Haojun and
Kai, Zhang and
Liu, Weize and
Chen, Danny and
Wu, Jian and
Chen, Jintai",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.517",
doi = "10.18653/v1/2023.findings-emnlp.517",
pages = "7705--7720",
abstract = "Deep learning approaches exhibit promising performances on various text tasks. However, they are still struggling on medical text classification since samples are often extremely imbalanced and scarce. Different from existing mainstream approaches that focus on supplementary semantics with external medical information, this paper aims to rethink the data challenges in medical texts and present a novel framework-agnostic algorithm called Text2Tree that only utilizes internal label hierarchy in training deep learning models. We embed the ICD code tree structure of labels into cascade attention modules for learning hierarchy-aware label representations. Two new learning schemes, Similarity Surrogate Learning (SSL) and Dissimilarity Mixup Learning (DML), are devised to boost text classification by reusing and distinguishing samples of other labels following the label representation hierarchy, respectively. Experiments on authoritative public datasets and real-world medical records show that our approach stably achieves superior performances over classical and advanced imbalanced classification methods. Our code is available at https://github.com/jyansir/Text2Tree.",
}
| Deep learning approaches exhibit promising performances on various text tasks. However, they are still struggling on medical text classification since samples are often extremely imbalanced and scarce. Different from existing mainstream approaches that focus on supplementary semantics with external medical information, this paper aims to rethink the data challenges in medical texts and present a novel framework-agnostic algorithm called Text2Tree that only utilizes internal label hierarchy in training deep learning models. We embed the ICD code tree structure of labels into cascade attention modules for learning hierarchy-aware label representations. Two new learning schemes, Similarity Surrogate Learning (SSL) and Dissimilarity Mixup Learning (DML), are devised to boost text classification by reusing and distinguishing samples of other labels following the label representation hierarchy, respectively. Experiments on authoritative public datasets and real-world medical records show that our approach stably achieves superior performances over classical and advanced imbalanced classification methods. Our code is available at https://github.com/jyansir/Text2Tree. | [
"Yan, Jiahuan",
"Gao, Haojun",
"Kai, Zhang",
"Liu, Weize",
"Chen, Danny",
"Wu, Jian",
"Chen, Jintai"
] | Text2Tree: Aligning Text Representation to the Label Tree Hierarchy for Imbalanced Medical Classification | findings-emnlp.517 | 2311.16650 | [
"https://github.com/jyansir/text2tree"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.518.bib | https://aclanthology.org/2023.findings-emnlp.518/ | @inproceedings{kang-choi-2023-impact,
title = "Impact of Co-occurrence on Factual Knowledge of Large Language Models",
author = "Kang, Cheongwoong and
Choi, Jaesik",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.518",
doi = "10.18653/v1/2023.findings-emnlp.518",
pages = "7721--7735",
abstract = "Large language models (LLMs) often make factually incorrect responses despite their success in various applications. In this paper, we hypothesize that relying heavily on simple co-occurrence statistics of the pre-training corpora is one of the main factors that cause factual errors. Our results reveal that LLMs are vulnerable to the co-occurrence bias, defined as preferring frequently co-occurred words over the correct answer. Consequently, LLMs struggle to recall facts whose subject and object rarely co-occur in the pre-training dataset although they are seen during finetuning. We show that co-occurrence bias remains despite scaling up model sizes or finetuning. Therefore, we suggest finetuning on a debiased dataset to mitigate the bias by filtering out biased samples whose subject-object co-occurrence count is high. Although debiased finetuning allows LLMs to memorize rare facts in the training set, it is not effective in recalling rare facts unseen during finetuning. Further research in mitigation will help build reliable language models by preventing potential errors. The code is available at https://github.com/CheongWoong/impact{\_}of{\_}cooccurrence.",
}
| Large language models (LLMs) often make factually incorrect responses despite their success in various applications. In this paper, we hypothesize that relying heavily on simple co-occurrence statistics of the pre-training corpora is one of the main factors that cause factual errors. Our results reveal that LLMs are vulnerable to the co-occurrence bias, defined as preferring frequently co-occurred words over the correct answer. Consequently, LLMs struggle to recall facts whose subject and object rarely co-occur in the pre-training dataset although they are seen during finetuning. We show that co-occurrence bias remains despite scaling up model sizes or finetuning. Therefore, we suggest finetuning on a debiased dataset to mitigate the bias by filtering out biased samples whose subject-object co-occurrence count is high. Although debiased finetuning allows LLMs to memorize rare facts in the training set, it is not effective in recalling rare facts unseen during finetuning. Further research in mitigation will help build reliable language models by preventing potential errors. The code is available at https://github.com/CheongWoong/impact{\_}of{\_}cooccurrence. | [
"Kang, Cheongwoong",
"Choi, Jaesik"
] | Impact of Co-occurrence on Factual Knowledge of Large Language Models | findings-emnlp.518 | 2310.08256 | [
"https://github.com/cheongwoong/impact_of_cooccurrence"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.findings-emnlp.519.bib | https://aclanthology.org/2023.findings-emnlp.519/ | @inproceedings{m-etal-2023-ctqscorer,
title = "{CTQS}corer: Combining Multiple Features for In-context Example Selection for Machine Translation",
author = "Kumar, Aswanth and
Puduppully, Ratish and
Dabre, Raj and
Kunchukuttan, Anoop",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-emnlp.519",
doi = "10.18653/v1/2023.findings-emnlp.519",
pages = "7736--7752",
abstract = "Large language models have demonstrated the capability to perform on machine translation when the input is prompted with a few examples (in-context learning). Translation quality depends on various features of the selected examples, such as their quality and relevance, but previous work has predominantly focused on individual features in isolation. In this paper, we propose a general framework for combining different features influencing example selection. We learn a regression model, CTQ Scorer (Contextual Translation Quality), that selects examples based on multiple features in order to maximize the translation quality. On multiple language pairs and language models, we show that CTQ Scorer helps significantly outperform random selection as well as strong single-factor baselines reported in the literature. We also see an improvement of over 2.5 COMET points on average with respect to a strong BM25 retrieval-based baseline.",
}
| Large language models have demonstrated the capability to perform on machine translation when the input is prompted with a few examples (in-context learning). Translation quality depends on various features of the selected examples, such as their quality and relevance, but previous work has predominantly focused on individual features in isolation. In this paper, we propose a general framework for combining different features influencing example selection. We learn a regression model, CTQ Scorer (Contextual Translation Quality), that selects examples based on multiple features in order to maximize the translation quality. On multiple language pairs and language models, we show that CTQ Scorer helps significantly outperform random selection as well as strong single-factor baselines reported in the literature. We also see an improvement of over 2.5 COMET points on average with respect to a strong BM25 retrieval-based baseline. | [
"Kumar, Aswanth",
"Puduppully, Ratish",
"Dabre, Raj",
"Kunchukuttan, Anoop"
] | CTQScorer: Combining Multiple Features for In-context Example Selection for Machine Translation | findings-emnlp.519 | 2305.14105 | [
"https://github.com/ai4bharat/ctqscorer"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.