init

.gitattributes

@@ -60,3 +60,4 @@ data/t_rex.raw.test.jsonl filter=lfs diff=lfs merge=lfs -text
 data/t_rex.filter.validation.jsonl filter=lfs diff=lfs merge=lfs -text
 data/t_rex.filter.test.jsonl filter=lfs diff=lfs merge=lfs -text
 data/t_rex.filter.train.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -22,19 +22,11 @@ We split the raw T-REX dataset into train/validation/test split by the ratio of
 
 We apply filtering to keep triples with alpha-numeric subject and object, as well as triples with at least either of subject or object is a named-entity.
 
-- Number of unique entities in subject and object.
-
-| Dataset | `train` | `validation` | `test`  | `all`   |
-|-------:|-------:|------------:|-------:|-------:|
-| raw     | 659,163 | 141,248      | 141,249 | 941,660 |
-| filter  | 463,521 | 99,550       | 99,408  | 662,479 |
-
-- Number of unique predicate.
-
-| Dataset | `train` | `validation` | `test`  | `all` |
-|-------:|-------:|------------:|-------:|-----:|
-| raw     | 894 | 717 | 700 | 2,311 | 
-| filter  | 780 | 614 | 616 | 2,010 |
+| Dataset | Raw | Filter |
+|--------:|----:|-------:|
+| Triples | 941,663 | 662,482 |
+| Predicate| 931 | 818 |
+| Entity | 270,801 | 197,302 |
 
 ### Filtering to Purify the Dataset
 We reduce the size of the dataset by applying filtering based on the number of predicates and entities in the triples.

data/{t_rex.filter.test.jsonl → t_rex.filter.jsonl}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f0cb4f880cdd24589aad630ba79e042a4dcce3d6471f490cd472886366a68d50
-size 90649913
+oid sha256:1b80224241efafb2b053bc630288a3adfbaf69422454b57c620c6f810da20aee
+size 603995857

data/t_rex.filter.validation.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b690964fd198d372882d889cc6169184be5373335315fc31e5e34975fb4d4c89
-size 90752371

data/{t_rex.filter.train.jsonl → t_rex.raw.jsonl}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:800dc2faef4a84c10be28a8560c5970a797e1ab852238367d0cec96f8db86021
-size 422593571
+oid sha256:f50215186e30015c38d374bdd737d989497852aacfb9ae3aeb467ec47537ad6a
+size 841348762

data/t_rex.raw.test.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:7255c9f4d2d5155e76075dcd5cee901e16903425e041ce76f7fc389265e89518
-size 125893572

data/t_rex.raw.train.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:9c664392be27e68c9a8172f80d76aa17eedac052b45623274afd8c2b060f2284
-size 589023853

data/t_rex.raw.validation.jsonl

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:44f9999f051074e8684cccc9c492718be8f3b36b55963eeede4d76323fa6c496
-size 126431335

filtering.py → filtering_denoise.py

@@ -38,11 +38,10 @@ def filtering(entry):
     return True
 
 
-for s in ['train', 'validation', 'test']:
-    with open(f"data/t_rex.raw.{s}.jsonl") as f:
-        data = [json.loads(i) for i in f.read().split('\n') if len(i) > 0]
-    print(f"[{s}] (before): {len(data)}")
-    data = [i for i in data if filtering(i)]
-    print(f"[{s}] (after) : {len(data)}")
-    with open(f"data/t_rex.filter.{s}.jsonl", 'w') as f:
-        f.write('\n'.join([json.dumps(i) for i in data]))
+with open(f"data/t_rex.raw.jsonl") as f:
+    data = [json.loads(i) for i in f.read().split('\n') if len(i) > 0]
+print(f"[before]: {len(data)}")
+data = [i for i in data if filtering(i)]
+print(f"[after] : {len(data)}")
+with open(f"data/t_rex.filter.jsonl", 'w') as f:
+    f.write('\n'.join([json.dumps(i) for i in data]))

stats.py → filtering_purify.py

@@ -15,13 +15,10 @@ from datasets import Dataset
 sns.set_theme(style="whitegrid")
 
 # load filtered data
-tmp = []
-splits = []
-for s in ['train', 'validation', 'test']:
-    with open(f"data/t_rex.filter.{s}.jsonl") as f:
-        _tmp = [json.loads(i) for i in f.read().split('\n') if len(i) > 0]
-        tmp += _tmp
-        splits += [s] * len(_tmp)
+with open(f"data/t_rex.filter.jsonl") as f:
+    _tmp = [json.loads(i) for i in f.read().split('\n') if len(i) > 0]
+    tmp += _tmp
+    splits += [s] * len(_tmp)
 data = Dataset.from_list(tmp)
 df_main = data.to_pandas()
 df_main['split'] = splits
@@ -92,7 +89,7 @@ if __name__ == '__main__':
     e_dist_full = []
     data_size_full = []
     config = []
-    candidates = list(product([1, 2, 3, 4], [100, 50, 25, 10]))
+    candidates = list(product([4, 8, 12, 16], [100, 50, 25, 10]))
 
     # run filtering with different configs
     for min_e_freq, max_p_freq in candidates:
@@ -102,9 +99,11 @@ if __name__ == '__main__':
         data_size_full.append(data_size)
         config.append([min_e_freq, max_p_freq])
         # save data
+        out = {}
         for s in ['train', 'validation', 'test']:
-            new_data_s = [i for i in new_data if i['split'] == s]
-            for i in new_data_s:
+            out[s] = [i for i in new_data if i['split'] == s]
+        for s, v in out.items():
+            for i in v:
                 i.pop('split')
             with open(f"data/t_rex.clean.min_entity_{min_e_freq}_max_predicate_{max_p_freq}.{s}.jsonl", 'w') as f:
                 f.write('\n'.join([json.dumps(i) for i in new_data_s]))

process.py

@@ -7,7 +7,7 @@ import json
 import os
 from glob import glob
 from tqdm import tqdm
-from random import shuffle, seed
+# from random import shuffle, seed
 
 os.makedirs('data', exist_ok=True)
 f_writer = open('data/t_rex.raw.jsonl', 'w')
@@ -25,13 +25,9 @@ for i in tqdm(glob("*.json")):
             f_writer.write(json.dumps(out) + "\n")
 f_writer.close()
 
-seed(0)
+
 with open('data/t_rex.raw.jsonl') as f:
     data = [json.loads(i) for i in f.read().split('\n') if len(i) > 0]
-shuffle(data)
-train = data[:int(len(data) * 0.7)]
-val = data[int(len(data) * 0.7):int(len(data) * 0.85)]
-test = data[int(len(data) * 0.85):]
-for i, j in zip([train, val, test], ['train', 'validation', 'test']):
-    with open(f'data/t_rex.raw.{j}.jsonl', 'w') as f:
-        f.write('\n'.join([json.dumps(l) for l in i]))
+s = {i['subject'] for i in data}
+o = {i['object'] for i in data}
+s.update(o)

t_rex.py

@@ -5,7 +5,7 @@ import datasets
 logger = datasets.logging.get_logger(__name__)
 _DESCRIPTION = """T-Rex dataset."""
 _NAME = "t_rex"
-_VERSION = "0.0.2"
+_VERSION = "0.0.3"
 _CITATION = """
 @inproceedings{elsahar2018t,
   title={T-rex: A large scale alignment of natural language with knowledge base triples},
@@ -18,11 +18,11 @@ _CITATION = """
 _HOME_PAGE = "https://github.com/asahi417/relbert"
 _URL = f'https://huggingface.co/datasets/relbert/{_NAME}/resolve/main/data'
 _TYPES = ["raw", "filter"]
-_URLS = {i: {
+_URLS = {i: {str(datasets.Split.TRAIN): [f'{_URL}/t_rex.{i}.jsonl']} if i in ['raw', 'filter'] else {
     str(datasets.Split.TRAIN): [f'{_URL}/t_rex.{i}.train.jsonl'],
     str(datasets.Split.VALIDATION): [f'{_URL}/t_rex.{i}.validation.jsonl'],
-    str(datasets.Split.TEST): [f'{_URL}/t_rex.{i}.test.jsonl']
-} for i in _TYPES}
+    str(datasets.Split.TEST): [f'{_URL}/t_rex.{i}.test.jsonl']}
+         for i in _TYPES}
 
 
 class TREXConfig(datasets.BuilderConfig):
@@ -46,8 +46,13 @@ class TREX(datasets.GeneratorBasedBuilder):
 
     def _split_generators(self, dl_manager):
         downloaded_file = dl_manager.download_and_extract(_URLS[self.config.name])
-        return [datasets.SplitGenerator(name=i, gen_kwargs={"filepaths": downloaded_file[str(i)]})
-                for i in [datasets.Split.TRAIN, datasets.Split.VALIDATION, datasets.Split.TEST]]
+        if self.config.name in ['raw', 'filter']:
+            return [datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={"filepaths": downloaded_file[str(datasets.Split.TRAIN)]})]
+        else:
+            return [datasets.SplitGenerator(name=i, gen_kwargs={"filepaths": downloaded_file[str(i)]})
+                    for i in [datasets.Split.TRAIN, datasets.Split.VALIDATION, datasets.Split.TEST]]
 
     def _generate_examples(self, filepaths):
         _key = 0