Upload globalwoz.py with huggingface_hub
Browse files- globalwoz.py +226 -0
globalwoz.py
ADDED
@@ -0,0 +1,226 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import json
|
3 |
+
from pathlib import Path
|
4 |
+
from typing import Dict, List, Tuple
|
5 |
+
|
6 |
+
import datasets
|
7 |
+
import itertools
|
8 |
+
|
9 |
+
from seacrowd.utils import schemas
|
10 |
+
from seacrowd.utils.configs import SEACrowdConfig
|
11 |
+
from seacrowd.utils.constants import Tasks, Licenses
|
12 |
+
|
13 |
+
_CITATION = """\
|
14 |
+
@inproceedings{ding-etal-2022-globalwoz,
|
15 |
+
title = "{G}lobal{W}o{Z}: Globalizing {M}ulti{W}o{Z} to Develop Multilingual Task-Oriented Dialogue Systems",
|
16 |
+
author = "Ding, Bosheng and
|
17 |
+
Hu, Junjie and
|
18 |
+
Bing, Lidong and
|
19 |
+
Aljunied, Mahani and
|
20 |
+
Joty, Shafiq and
|
21 |
+
Si, Luo and
|
22 |
+
Miao, Chunyan",
|
23 |
+
editor = "Muresan, Smaranda and
|
24 |
+
Nakov, Preslav and
|
25 |
+
Villavicencio, Aline",
|
26 |
+
booktitle = "Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
|
27 |
+
month = may,
|
28 |
+
year = "2022",
|
29 |
+
}
|
30 |
+
"""
|
31 |
+
|
32 |
+
_DATASETNAME = "globalwoz"
|
33 |
+
|
34 |
+
_DESCRIPTION = """\
|
35 |
+
This is the data of the paper “GlobalWoZ: Globalizing MultiWoZ to Develop Multilingual Task-Oriented Dialogue Systems” accepted by ACL 2022. The dataset contains several sub-datasets in 20 languages and 3 schemes (F&E, E&F, F&F), including Indonesian (id), Thai (th), and Vietnamese (vi) language. The method is based on translating dialogue templates and filling them with local entities in the target language countries.
|
36 |
+
"""
|
37 |
+
|
38 |
+
|
39 |
+
_HOMEPAGE = "https://github.com/bosheng2020/globalwoz"
|
40 |
+
|
41 |
+
|
42 |
+
_LANGUAGES = ["ind", "tha", "vie"]
|
43 |
+
|
44 |
+
_LICENSE = Licenses.UNKNOWN.value
|
45 |
+
|
46 |
+
_LOCAL = True
|
47 |
+
|
48 |
+
_URLS = {}
|
49 |
+
|
50 |
+
_SUPPORTED_TASKS = [Tasks.E2E_TASK_ORIENTED_DIALOGUE]
|
51 |
+
|
52 |
+
_SOURCE_VERSION = "2.0.0"
|
53 |
+
|
54 |
+
_SEACROWD_VERSION = "2024.06.20"
|
55 |
+
|
56 |
+
|
57 |
+
def seacrowd_config_constructor(dial_type, lang, schema, version):
|
58 |
+
if dial_type not in ["EandF", "FandE", "FandF"]:
|
59 |
+
raise ValueError(f"Invalid dialogue type {dial_type}")
|
60 |
+
|
61 |
+
if lang == "":
|
62 |
+
raise ValueError(f"Invalid lang {lang}")
|
63 |
+
|
64 |
+
if schema not in ["source", "seacrowd_tod"]:
|
65 |
+
raise ValueError(f"Invalid schema: {schema}")
|
66 |
+
|
67 |
+
return SEACrowdConfig(
|
68 |
+
name="globalwoz_{dial_type}_{lang}_{schema}".format(dial_type=dial_type, lang=lang, schema=schema),
|
69 |
+
version=datasets.Version(version),
|
70 |
+
description="GlobalWoZ schema for {schema}: {dial_type}_{lang}".format(schema=schema, dial_type=dial_type, lang=lang),
|
71 |
+
schema=schema,
|
72 |
+
subset_id="globalwoz_{dial_type}_{lang}".format(dial_type=dial_type, lang=lang),
|
73 |
+
)
|
74 |
+
|
75 |
+
|
76 |
+
class GlobalWoZ(datasets.GeneratorBasedBuilder):
|
77 |
+
"""This is the data of the paper “GlobalWoZ: Globalizing MultiWoZ to Develop Multilingual Task-Oriented Dialogue Systems” accepted by ACL 2022.
|
78 |
+
The dataset contains several sub-datasets in 20 languages and 3 schemes (F&E, E&F, F&F), including Indonesian (id), Thai (th),
|
79 |
+
and Vietnamese (vi) language. The method is based on translating dialogue templates and filling them with local entities in the target language countries.
|
80 |
+
"""
|
81 |
+
|
82 |
+
SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
|
83 |
+
SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION)
|
84 |
+
|
85 |
+
BUILDER_CONFIGS = [
|
86 |
+
seacrowd_config_constructor(tod_format, lang, schema, _SOURCE_VERSION if schema == "source" else _SEACROWD_VERSION) for tod_format, lang, schema in itertools.product(("EandF", "FandE", "FandF"), ("id", "th", "vi"), ("source", "seacrowd_tod"))
|
87 |
+
]
|
88 |
+
|
89 |
+
def _info(self) -> datasets.DatasetInfo:
|
90 |
+
if self.config.schema == "source":
|
91 |
+
features = datasets.Features(
|
92 |
+
{
|
93 |
+
"id": datasets.Value("string"),
|
94 |
+
"goal": {
|
95 |
+
"attraction": datasets.Value("string"),
|
96 |
+
"hospital": datasets.Value("string"),
|
97 |
+
"hotel": datasets.Value("string"),
|
98 |
+
"police": datasets.Value("string"),
|
99 |
+
"restaurant": datasets.Value("string"),
|
100 |
+
"taxi": datasets.Value("string"),
|
101 |
+
"train": datasets.Value("string"),
|
102 |
+
},
|
103 |
+
"log": [
|
104 |
+
{
|
105 |
+
"dialog_act": datasets.Value("string"),
|
106 |
+
"metadata": datasets.Value("string"),
|
107 |
+
"span_info": [[datasets.Value("string")]],
|
108 |
+
"text": datasets.Value("string"),
|
109 |
+
}
|
110 |
+
],
|
111 |
+
}
|
112 |
+
)
|
113 |
+
|
114 |
+
elif self.config.schema == "seacrowd_tod":
|
115 |
+
features = schemas.tod_features
|
116 |
+
else:
|
117 |
+
raise NotImplementedError()
|
118 |
+
|
119 |
+
return datasets.DatasetInfo(
|
120 |
+
description=_DESCRIPTION,
|
121 |
+
features=features,
|
122 |
+
homepage=_HOMEPAGE,
|
123 |
+
license=_LICENSE,
|
124 |
+
citation=_CITATION,
|
125 |
+
)
|
126 |
+
|
127 |
+
def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
|
128 |
+
"""Returns SplitGenerators."""
|
129 |
+
_split_generators = []
|
130 |
+
|
131 |
+
type_and_lang = {"dial_type": self.config.subset_id.split("_")[1].replace("and", "&"), "lang": self.config.subset_id.split("_")[2]} # globalwoz_{dial_type}_{lang}
|
132 |
+
|
133 |
+
if self.config.data_dir is None:
|
134 |
+
raise ValueError("This is a local dataset. Please pass the data_dir kwarg to load_dataset.")
|
135 |
+
else:
|
136 |
+
data_dir = self.config.data_dir
|
137 |
+
|
138 |
+
if not os.path.exists(os.path.join(data_dir, f"{type_and_lang['dial_type']}_{type_and_lang['lang']}.json")):
|
139 |
+
raise FileNotFoundError()
|
140 |
+
|
141 |
+
return [
|
142 |
+
datasets.SplitGenerator(
|
143 |
+
name=datasets.Split.TRAIN,
|
144 |
+
gen_kwargs={
|
145 |
+
# "filepath": data_dir + f"_{type_and_lang['dial_type']}_{type_and_lang['lang']}.json",
|
146 |
+
"filepath": os.path.join(data_dir, f"{type_and_lang['dial_type']}_{type_and_lang['lang']}.json"),
|
147 |
+
"split": "train",
|
148 |
+
},
|
149 |
+
),
|
150 |
+
]
|
151 |
+
|
152 |
+
def _generate_examples(self, filepath: Path, split: str) -> Tuple[int, Dict]:
|
153 |
+
"""Yields examples as (key, example) tuples."""
|
154 |
+
# For local datasets you will have access to self.config.data_dir and self.config.data_files
|
155 |
+
with open(filepath, "r+", encoding="utf8") as fw:
|
156 |
+
data = json.load(fw)
|
157 |
+
|
158 |
+
if self.config.schema == "source":
|
159 |
+
for idx, tod_dialogue in enumerate(data.values()):
|
160 |
+
example = {}
|
161 |
+
example["id"] = str(idx)
|
162 |
+
example["goal"] = {}
|
163 |
+
|
164 |
+
for goal_key in ["attraction", "hospital", "hotel", "police", "restaurant", "taxi", "train"]:
|
165 |
+
example["goal"][goal_key] = json.dumps(tod_dialogue["goal"][goal_key])
|
166 |
+
example["log"] = []
|
167 |
+
|
168 |
+
for dial_log in tod_dialogue["log"]:
|
169 |
+
dial = {}
|
170 |
+
dial["dialog_act"] = json.dumps(dial_log["dialog_act"])
|
171 |
+
dial["metadata"] = json.dumps(dial_log["metadata"])
|
172 |
+
for i in range(len(dial_log["span_info"])):
|
173 |
+
for j in range(len(dial_log["span_info"][i])):
|
174 |
+
dial_log["span_info"][i][j] = str(dial_log["span_info"][i][j]) # casting to str
|
175 |
+
dial["span_info"] = [[str(span)] if isinstance(span, str) else span for span in dial_log["span_info"]]
|
176 |
+
dial["text"] = dial_log["text"]
|
177 |
+
|
178 |
+
example["log"].append(dial)
|
179 |
+
|
180 |
+
yield example["id"], example
|
181 |
+
|
182 |
+
elif self.config.schema == "seacrowd_tod":
|
183 |
+
for idx, tod_dialogue in enumerate(data.values()):
|
184 |
+
example = {}
|
185 |
+
example["dialogue_idx"] = idx
|
186 |
+
|
187 |
+
dialogue = []
|
188 |
+
# NOTE: the dialogue always started with `user` as first utterance
|
189 |
+
for turn, i in enumerate(range(0, len(tod_dialogue["log"]) + 2, 2)):
|
190 |
+
dial = {}
|
191 |
+
dial["turn_idx"] = turn
|
192 |
+
|
193 |
+
# system_utterance properties
|
194 |
+
dial["system_utterance"] = ""
|
195 |
+
dial["system_acts"] = []
|
196 |
+
if turn != 0:
|
197 |
+
dial["system_utterance"] = tod_dialogue["log"][i - 1]["text"]
|
198 |
+
if i < len(tod_dialogue["log"]):
|
199 |
+
# NOTE: "system_acts will be populated with the `dialog_act` from the user utterance in the original dataset, as our schema dictates
|
200 |
+
# that `system_acts` should represent the system's intended actions based on the user's utterance."
|
201 |
+
for acts in tod_dialogue["log"][i]["dialog_act"].values():
|
202 |
+
for act in acts:
|
203 |
+
dial["system_acts"].append([act[0]])
|
204 |
+
|
205 |
+
# user_utterance properties
|
206 |
+
dial["turn_label"] = [] # left as an empty array
|
207 |
+
dial["belief_state"] = []
|
208 |
+
if i == len(tod_dialogue["log"]):
|
209 |
+
# case if turn_idx > len(dialogue) --> add dummy user_utterance
|
210 |
+
dial["user_utterance"] = ""
|
211 |
+
else:
|
212 |
+
dial["user_utterance"] = tod_dialogue["log"][i]["text"]
|
213 |
+
# NOTE: "the belief_state will be populated with the `span_info` from the user utterance in the original dataset, as our schema dictates
|
214 |
+
# that `belief_state` should represent the system's belief state based on the user's utterance."
|
215 |
+
for span in tod_dialogue["log"][i]["span_info"]:
|
216 |
+
if span[0].split("-")[1] == "request": # Request action
|
217 |
+
dial["belief_state"].append({"slots": [["slot", span[1]]], "act": "request"})
|
218 |
+
else:
|
219 |
+
dial["belief_state"].append({"slots": [[span[1], span[2]]], "act": span[0].split("-")[1]})
|
220 |
+
|
221 |
+
# append to dialogue
|
222 |
+
dialogue.append(dial)
|
223 |
+
|
224 |
+
example["dialogue"] = dialogue
|
225 |
+
|
226 |
+
yield example["dialogue_idx"], example
|