Datasets:
Quentin Lhoest
commited on
Commit
·
e4c37ba
1
Parent(s):
99c0835
Release: 1.18.1
Browse filesCommit from https://github.com/huggingface/datasets/commit/218e496519ff14b4bc69ea559616af6f2ef89e57
lama.py
CHANGED
@@ -1,350 +1,350 @@
|
|
1 |
-
# coding=utf-8
|
2 |
-
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
|
3 |
-
#
|
4 |
-
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
-
# you may not use this file except in compliance with the License.
|
6 |
-
# You may obtain a copy of the License at
|
7 |
-
#
|
8 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
-
#
|
10 |
-
# Unless required by applicable law or agreed to in writing, software
|
11 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
-
# See the License for the specific language governing permissions and
|
14 |
-
# limitations under the License.
|
15 |
-
"""The LAMA Dataset"""
|
16 |
-
|
17 |
-
|
18 |
-
import json
|
19 |
-
from fnmatch import fnmatch
|
20 |
-
|
21 |
-
import datasets
|
22 |
-
|
23 |
-
|
24 |
-
_CITATION = """@inproceedings{petroni2019language,
|
25 |
-
title={Language Models as Knowledge Bases?},
|
26 |
-
author={F. Petroni, T. Rockt{\"{a}}schel, A. H. Miller, P. Lewis, A. Bakhtin, Y. Wu and S. Riedel},
|
27 |
-
booktitle={In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2019},
|
28 |
-
year={2019}
|
29 |
-
}
|
30 |
-
@inproceedings{petroni2020how,
|
31 |
-
title={How Context Affects Language Models' Factual Predictions},
|
32 |
-
author={Fabio Petroni and Patrick Lewis and Aleksandra Piktus and Tim Rockt{\"a}schel and Yuxiang Wu and Alexander H. Miller and Sebastian Riedel},
|
33 |
-
booktitle={Automated Knowledge Base Construction},
|
34 |
-
year={2020},
|
35 |
-
url={https://openreview.net/forum?id=025X0zPfn}
|
36 |
-
}
|
37 |
-
"""
|
38 |
-
|
39 |
-
|
40 |
-
_DESCRIPTION = """LAMA is a dataset used to probe and analyze the factual and commonsense knowledge contained in pretrained language models. See https://github.com/facebookresearch/LAMA.
|
41 |
-
"""
|
42 |
-
|
43 |
-
_HOMEPAGE = "https://github.com/facebookresearch/LAMA"
|
44 |
-
|
45 |
-
_LICENSE = "The Creative Commons Attribution-Noncommercial 4.0 International License. see https://github.com/facebookresearch/LAMA/blob/master/LICENSE"
|
46 |
-
|
47 |
-
_RELATIONS_URL = "https://s3.amazonaws.com/datasets.huggingface.co/lama/relations.jsonl"
|
48 |
-
|
49 |
-
_DATA_URL = "https://dl.fbaipublicfiles.com/LAMA/negated_data.tar.gz"
|
50 |
-
|
51 |
-
|
52 |
-
class Lama(datasets.GeneratorBasedBuilder):
|
53 |
-
"""Lama Dataset"""
|
54 |
-
|
55 |
-
VERSION = datasets.Version("1.1.0")
|
56 |
-
|
57 |
-
BUILDER_CONFIGS = [
|
58 |
-
datasets.BuilderConfig(name="trex", version=VERSION, description="The TRex part of the Lama dataset"),
|
59 |
-
datasets.BuilderConfig(name="squad", version=VERSION, description="The Squad part of the Lama dataset"),
|
60 |
-
datasets.BuilderConfig(
|
61 |
-
name="google_re", version=VERSION, description="The Google_re part of the Lama dataset"
|
62 |
-
),
|
63 |
-
datasets.BuilderConfig(
|
64 |
-
name="conceptnet", version=VERSION, description="The Conceptnet part of the Lama dataset"
|
65 |
-
),
|
66 |
-
]
|
67 |
-
|
68 |
-
DEFAULT_CONFIG_NAME = "trex"
|
69 |
-
|
70 |
-
def _info(self):
|
71 |
-
if self.config.name == "trex":
|
72 |
-
features = datasets.Features(
|
73 |
-
{
|
74 |
-
"uuid": datasets.Value("string"),
|
75 |
-
"obj_uri": datasets.Value("string"),
|
76 |
-
"obj_label": datasets.Value("string"),
|
77 |
-
"sub_uri": datasets.Value("string"),
|
78 |
-
"sub_label": datasets.Value("string"),
|
79 |
-
"predicate_id": datasets.Value("string"),
|
80 |
-
"sub_surface": datasets.Value("string"),
|
81 |
-
"obj_surface": datasets.Value("string"),
|
82 |
-
"masked_sentence": datasets.Value("string"),
|
83 |
-
"template": datasets.Value("string"),
|
84 |
-
"template_negated": datasets.Value("string"),
|
85 |
-
"label": datasets.Value("string"),
|
86 |
-
"description": datasets.Value("string"),
|
87 |
-
"type": datasets.Value("string"),
|
88 |
-
}
|
89 |
-
)
|
90 |
-
return datasets.DatasetInfo(
|
91 |
-
description=_DESCRIPTION,
|
92 |
-
features=features,
|
93 |
-
supervised_keys=None,
|
94 |
-
homepage=_HOMEPAGE,
|
95 |
-
license=_LICENSE,
|
96 |
-
citation=_CITATION,
|
97 |
-
)
|
98 |
-
elif self.config.name == "conceptnet":
|
99 |
-
features = datasets.Features(
|
100 |
-
{
|
101 |
-
"uuid": datasets.Value("string"),
|
102 |
-
"sub": datasets.Value("string"),
|
103 |
-
"obj": datasets.Value("string"),
|
104 |
-
"pred": datasets.Value("string"),
|
105 |
-
"obj_label": datasets.Value("string"),
|
106 |
-
"masked_sentence": datasets.Value("string"),
|
107 |
-
"negated": datasets.Value("string"),
|
108 |
-
}
|
109 |
-
)
|
110 |
-
return datasets.DatasetInfo(
|
111 |
-
description=_DESCRIPTION,
|
112 |
-
features=features,
|
113 |
-
supervised_keys=None,
|
114 |
-
homepage=_HOMEPAGE,
|
115 |
-
license=_LICENSE,
|
116 |
-
citation=_CITATION,
|
117 |
-
)
|
118 |
-
elif self.config.name == "squad":
|
119 |
-
features = datasets.Features(
|
120 |
-
{
|
121 |
-
"id": datasets.Value("string"),
|
122 |
-
"sub_label": datasets.Value("string"),
|
123 |
-
"obj_label": datasets.Value("string"),
|
124 |
-
"negated": datasets.Value("string"),
|
125 |
-
"masked_sentence": datasets.Value("string"),
|
126 |
-
}
|
127 |
-
)
|
128 |
-
return datasets.DatasetInfo(
|
129 |
-
description=_DESCRIPTION,
|
130 |
-
features=features,
|
131 |
-
supervised_keys=None,
|
132 |
-
homepage=_HOMEPAGE,
|
133 |
-
license=_LICENSE,
|
134 |
-
citation=_CITATION,
|
135 |
-
)
|
136 |
-
elif self.config.name == "google_re":
|
137 |
-
features = datasets.Features(
|
138 |
-
{
|
139 |
-
"pred": datasets.Value("string"),
|
140 |
-
"sub": datasets.Value("string"),
|
141 |
-
"obj": datasets.Value("string"),
|
142 |
-
"evidences": datasets.Value("string"),
|
143 |
-
"judgments": datasets.Value("string"),
|
144 |
-
"sub_w": datasets.Value("string"),
|
145 |
-
"sub_label": datasets.Value("string"),
|
146 |
-
"sub_aliases": datasets.Value("string"),
|
147 |
-
"obj_w": datasets.Value("string"),
|
148 |
-
"obj_label": datasets.Value("string"),
|
149 |
-
"obj_aliases": datasets.Value("string"),
|
150 |
-
"uuid": datasets.Value("string"),
|
151 |
-
"masked_sentence": datasets.Value("string"),
|
152 |
-
"template": datasets.Value("string"),
|
153 |
-
"template_negated": datasets.Value("string"),
|
154 |
-
}
|
155 |
-
)
|
156 |
-
return datasets.DatasetInfo(
|
157 |
-
description=_DESCRIPTION,
|
158 |
-
features=features,
|
159 |
-
supervised_keys=None,
|
160 |
-
homepage=_HOMEPAGE,
|
161 |
-
license=_LICENSE,
|
162 |
-
citation=_CITATION,
|
163 |
-
)
|
164 |
-
|
165 |
-
def _split_generators(self, dl_manager):
|
166 |
-
"""Returns SplitGenerators."""
|
167 |
-
archive = dl_manager.download(_DATA_URL)
|
168 |
-
if self.config.name == "trex":
|
169 |
-
relations_path = dl_manager.download(_RELATIONS_URL)
|
170 |
-
return [
|
171 |
-
datasets.SplitGenerator(
|
172 |
-
name=datasets.Split.TRAIN,
|
173 |
-
gen_kwargs={
|
174 |
-
"filepaths": ["TREx/*"],
|
175 |
-
"files": dl_manager.iter_archive(archive),
|
176 |
-
"relations_path": relations_path,
|
177 |
-
},
|
178 |
-
),
|
179 |
-
]
|
180 |
-
elif self.config.name == "google_re":
|
181 |
-
return [
|
182 |
-
datasets.SplitGenerator(
|
183 |
-
name=datasets.Split.TRAIN,
|
184 |
-
gen_kwargs={
|
185 |
-
"filepaths": [
|
186 |
-
"Google_RE/date_of_birth_test.jsonl",
|
187 |
-
"Google_RE/place_of_birth_test.jsonl",
|
188 |
-
"Google_RE/place_of_death_test.jsonl",
|
189 |
-
],
|
190 |
-
"files": dl_manager.iter_archive(archive),
|
191 |
-
},
|
192 |
-
),
|
193 |
-
]
|
194 |
-
elif self.config.name == "conceptnet":
|
195 |
-
return [
|
196 |
-
datasets.SplitGenerator(
|
197 |
-
name=datasets.Split.TRAIN,
|
198 |
-
gen_kwargs={
|
199 |
-
"filepaths": ["ConceptNet/test.jsonl"],
|
200 |
-
"files": dl_manager.iter_archive(archive),
|
201 |
-
},
|
202 |
-
),
|
203 |
-
]
|
204 |
-
elif self.config.name == "squad":
|
205 |
-
return [
|
206 |
-
datasets.SplitGenerator(
|
207 |
-
name=datasets.Split.TRAIN,
|
208 |
-
gen_kwargs={
|
209 |
-
"filepaths": ["Squad/test.jsonl"],
|
210 |
-
"files": dl_manager.iter_archive(archive),
|
211 |
-
},
|
212 |
-
),
|
213 |
-
]
|
214 |
-
|
215 |
-
def _generate_examples(self, filepaths, files, relations_path=None):
|
216 |
-
"""Yields examples from the LAMA dataset."""
|
217 |
-
filepaths = list(filepaths)
|
218 |
-
if self.config.name == "trex":
|
219 |
-
all_rels = {}
|
220 |
-
with open(relations_path, encoding="utf-8") as f:
|
221 |
-
for row in f:
|
222 |
-
data = json.loads(row)
|
223 |
-
all_rels[data["relation"]] = data
|
224 |
-
id_ = -1
|
225 |
-
inside_trec_directory = False
|
226 |
-
for path, f in files:
|
227 |
-
if any(fnmatch(path, pattern) for pattern in filepaths):
|
228 |
-
inside_trec_directory = True
|
229 |
-
for row in f:
|
230 |
-
data = json.loads(row)
|
231 |
-
pred = all_rels.get(data["predicate_id"], {})
|
232 |
-
for evidences in data["evidences"]:
|
233 |
-
id_ += 1
|
234 |
-
yield id_, {
|
235 |
-
"uuid": str(data["uuid"]),
|
236 |
-
"obj_uri": str(data["obj_uri"]),
|
237 |
-
"obj_label": str(data["obj_label"]),
|
238 |
-
"sub_uri": str(data["sub_uri"]),
|
239 |
-
"sub_label": str(data["sub_label"]),
|
240 |
-
"predicate_id": str(data["predicate_id"]),
|
241 |
-
"sub_surface": str(evidences["sub_surface"]),
|
242 |
-
"obj_surface": str(evidences["obj_surface"]),
|
243 |
-
"masked_sentence": str(evidences["masked_sentence"]),
|
244 |
-
"template": str(pred.get("template", "")),
|
245 |
-
"template_negated": str(pred.get("template_negated", "")),
|
246 |
-
"label": str(pred.get("label", "")),
|
247 |
-
"description": str(pred.get("description", "")),
|
248 |
-
"type": str(pred.get("type", "")),
|
249 |
-
}
|
250 |
-
elif inside_trec_directory:
|
251 |
-
break
|
252 |
-
elif self.config.name == "conceptnet":
|
253 |
-
id_ = -1
|
254 |
-
for path, f in files:
|
255 |
-
if not filepaths:
|
256 |
-
break
|
257 |
-
if path in list(filepaths):
|
258 |
-
for row in f:
|
259 |
-
data = json.loads(row)
|
260 |
-
if data.get("negated") is not None:
|
261 |
-
for masked_sentence, negated in zip(data["masked_sentences"], data["negated"]):
|
262 |
-
id_ += 1
|
263 |
-
yield id_, {
|
264 |
-
"uuid": str(data["uuid"]),
|
265 |
-
"sub": str(data.get("sub", "")),
|
266 |
-
"obj": str(data.get("obj", "")),
|
267 |
-
"pred": str(data["pred"]),
|
268 |
-
"obj_label": str(data["obj_label"]),
|
269 |
-
"masked_sentence": str(masked_sentence),
|
270 |
-
"negated": str(negated),
|
271 |
-
}
|
272 |
-
else:
|
273 |
-
for masked_sentence in data["masked_sentences"]:
|
274 |
-
id_ += 1
|
275 |
-
yield id_, {
|
276 |
-
"uuid": str(data["uuid"]),
|
277 |
-
"sub": str(data.get("sub", "")),
|
278 |
-
"obj": str(data.get("obj", "")),
|
279 |
-
"pred": str(data["pred"]),
|
280 |
-
"obj_label": str(data["obj_label"]),
|
281 |
-
"masked_sentence": str(masked_sentence),
|
282 |
-
"negated": str(""),
|
283 |
-
}
|
284 |
-
filepaths.remove(path)
|
285 |
-
elif self.config.name == "squad":
|
286 |
-
id_ = -1
|
287 |
-
for path, f in files:
|
288 |
-
if not filepaths:
|
289 |
-
break
|
290 |
-
if path in filepaths:
|
291 |
-
for row in f:
|
292 |
-
data = json.loads(row)
|
293 |
-
for masked_sentence in data["masked_sentences"]:
|
294 |
-
id_ += 1
|
295 |
-
yield id_, {
|
296 |
-
"id": str(data["id"]),
|
297 |
-
"sub_label": str(data["sub_label"]),
|
298 |
-
"obj_label": str(data["obj_label"]),
|
299 |
-
"negated": str(data.get("negated", "")),
|
300 |
-
"masked_sentence": str(masked_sentence),
|
301 |
-
}
|
302 |
-
filepaths.remove(path)
|
303 |
-
elif self.config.name == "google_re":
|
304 |
-
id_ = -1
|
305 |
-
for path, f in files:
|
306 |
-
if path in filepaths:
|
307 |
-
if not filepaths:
|
308 |
-
break
|
309 |
-
if path in filepaths:
|
310 |
-
# from https://github.com/facebookresearch/LAMA/blob/master/scripts/run_experiments.py
|
311 |
-
if "place_of_birth" in path:
|
312 |
-
pred = {
|
313 |
-
"relation": "place_of_birth",
|
314 |
-
"template": "[X] was born in [Y] .",
|
315 |
-
"template_negated": "[X] was not born in [Y] .",
|
316 |
-
}
|
317 |
-
elif "date_of_birth" in path:
|
318 |
-
pred = {
|
319 |
-
"relation": "date_of_birth",
|
320 |
-
"template": "[X] (born [Y]).",
|
321 |
-
"template_negated": "[X] (not born [Y]).",
|
322 |
-
}
|
323 |
-
else:
|
324 |
-
pred = {
|
325 |
-
"relation": "place_of_death",
|
326 |
-
"template": "[X] died in [Y] .",
|
327 |
-
"template_negated": "[X] did not die in [Y] .",
|
328 |
-
}
|
329 |
-
for row in f:
|
330 |
-
data = json.loads(row)
|
331 |
-
for masked_sentence in data["masked_sentences"]:
|
332 |
-
id_ += 1
|
333 |
-
yield id_, {
|
334 |
-
"pred": str(data["pred"]),
|
335 |
-
"sub": str(data["sub"]),
|
336 |
-
"obj": str(data["obj"]),
|
337 |
-
"evidences": str(data["evidences"]),
|
338 |
-
"judgments": str(data["judgments"]),
|
339 |
-
"sub_w": str(data["sub_w"]),
|
340 |
-
"sub_label": str(data["sub_label"]),
|
341 |
-
"sub_aliases": str(data["sub_aliases"]),
|
342 |
-
"obj_w": str(data["obj_w"]),
|
343 |
-
"obj_label": str(data["obj_label"]),
|
344 |
-
"obj_aliases": str(data["obj_aliases"]),
|
345 |
-
"uuid": str(data["uuid"]),
|
346 |
-
"masked_sentence": str(masked_sentence),
|
347 |
-
"template": str(pred["template"]),
|
348 |
-
"template_negated": str(pred["template_negated"]),
|
349 |
-
}
|
350 |
-
filepaths.remove(path)
|
|
|
1 |
+
# coding=utf-8
|
2 |
+
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
|
3 |
+
#
|
4 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
+
# you may not use this file except in compliance with the License.
|
6 |
+
# You may obtain a copy of the License at
|
7 |
+
#
|
8 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
+
#
|
10 |
+
# Unless required by applicable law or agreed to in writing, software
|
11 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
+
# See the License for the specific language governing permissions and
|
14 |
+
# limitations under the License.
|
15 |
+
"""The LAMA Dataset"""
|
16 |
+
|
17 |
+
|
18 |
+
import json
|
19 |
+
from fnmatch import fnmatch
|
20 |
+
|
21 |
+
import datasets
|
22 |
+
|
23 |
+
|
24 |
+
_CITATION = """@inproceedings{petroni2019language,
|
25 |
+
title={Language Models as Knowledge Bases?},
|
26 |
+
author={F. Petroni, T. Rockt{\"{a}}schel, A. H. Miller, P. Lewis, A. Bakhtin, Y. Wu and S. Riedel},
|
27 |
+
booktitle={In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2019},
|
28 |
+
year={2019}
|
29 |
+
}
|
30 |
+
@inproceedings{petroni2020how,
|
31 |
+
title={How Context Affects Language Models' Factual Predictions},
|
32 |
+
author={Fabio Petroni and Patrick Lewis and Aleksandra Piktus and Tim Rockt{\"a}schel and Yuxiang Wu and Alexander H. Miller and Sebastian Riedel},
|
33 |
+
booktitle={Automated Knowledge Base Construction},
|
34 |
+
year={2020},
|
35 |
+
url={https://openreview.net/forum?id=025X0zPfn}
|
36 |
+
}
|
37 |
+
"""
|
38 |
+
|
39 |
+
|
40 |
+
_DESCRIPTION = """LAMA is a dataset used to probe and analyze the factual and commonsense knowledge contained in pretrained language models. See https://github.com/facebookresearch/LAMA.
|
41 |
+
"""
|
42 |
+
|
43 |
+
_HOMEPAGE = "https://github.com/facebookresearch/LAMA"
|
44 |
+
|
45 |
+
_LICENSE = "The Creative Commons Attribution-Noncommercial 4.0 International License. see https://github.com/facebookresearch/LAMA/blob/master/LICENSE"
|
46 |
+
|
47 |
+
_RELATIONS_URL = "https://s3.amazonaws.com/datasets.huggingface.co/lama/relations.jsonl"
|
48 |
+
|
49 |
+
_DATA_URL = "https://dl.fbaipublicfiles.com/LAMA/negated_data.tar.gz"
|
50 |
+
|
51 |
+
|
52 |
+
class Lama(datasets.GeneratorBasedBuilder):
|
53 |
+
"""Lama Dataset"""
|
54 |
+
|
55 |
+
VERSION = datasets.Version("1.1.0")
|
56 |
+
|
57 |
+
BUILDER_CONFIGS = [
|
58 |
+
datasets.BuilderConfig(name="trex", version=VERSION, description="The TRex part of the Lama dataset"),
|
59 |
+
datasets.BuilderConfig(name="squad", version=VERSION, description="The Squad part of the Lama dataset"),
|
60 |
+
datasets.BuilderConfig(
|
61 |
+
name="google_re", version=VERSION, description="The Google_re part of the Lama dataset"
|
62 |
+
),
|
63 |
+
datasets.BuilderConfig(
|
64 |
+
name="conceptnet", version=VERSION, description="The Conceptnet part of the Lama dataset"
|
65 |
+
),
|
66 |
+
]
|
67 |
+
|
68 |
+
DEFAULT_CONFIG_NAME = "trex"
|
69 |
+
|
70 |
+
def _info(self):
|
71 |
+
if self.config.name == "trex":
|
72 |
+
features = datasets.Features(
|
73 |
+
{
|
74 |
+
"uuid": datasets.Value("string"),
|
75 |
+
"obj_uri": datasets.Value("string"),
|
76 |
+
"obj_label": datasets.Value("string"),
|
77 |
+
"sub_uri": datasets.Value("string"),
|
78 |
+
"sub_label": datasets.Value("string"),
|
79 |
+
"predicate_id": datasets.Value("string"),
|
80 |
+
"sub_surface": datasets.Value("string"),
|
81 |
+
"obj_surface": datasets.Value("string"),
|
82 |
+
"masked_sentence": datasets.Value("string"),
|
83 |
+
"template": datasets.Value("string"),
|
84 |
+
"template_negated": datasets.Value("string"),
|
85 |
+
"label": datasets.Value("string"),
|
86 |
+
"description": datasets.Value("string"),
|
87 |
+
"type": datasets.Value("string"),
|
88 |
+
}
|
89 |
+
)
|
90 |
+
return datasets.DatasetInfo(
|
91 |
+
description=_DESCRIPTION,
|
92 |
+
features=features,
|
93 |
+
supervised_keys=None,
|
94 |
+
homepage=_HOMEPAGE,
|
95 |
+
license=_LICENSE,
|
96 |
+
citation=_CITATION,
|
97 |
+
)
|
98 |
+
elif self.config.name == "conceptnet":
|
99 |
+
features = datasets.Features(
|
100 |
+
{
|
101 |
+
"uuid": datasets.Value("string"),
|
102 |
+
"sub": datasets.Value("string"),
|
103 |
+
"obj": datasets.Value("string"),
|
104 |
+
"pred": datasets.Value("string"),
|
105 |
+
"obj_label": datasets.Value("string"),
|
106 |
+
"masked_sentence": datasets.Value("string"),
|
107 |
+
"negated": datasets.Value("string"),
|
108 |
+
}
|
109 |
+
)
|
110 |
+
return datasets.DatasetInfo(
|
111 |
+
description=_DESCRIPTION,
|
112 |
+
features=features,
|
113 |
+
supervised_keys=None,
|
114 |
+
homepage=_HOMEPAGE,
|
115 |
+
license=_LICENSE,
|
116 |
+
citation=_CITATION,
|
117 |
+
)
|
118 |
+
elif self.config.name == "squad":
|
119 |
+
features = datasets.Features(
|
120 |
+
{
|
121 |
+
"id": datasets.Value("string"),
|
122 |
+
"sub_label": datasets.Value("string"),
|
123 |
+
"obj_label": datasets.Value("string"),
|
124 |
+
"negated": datasets.Value("string"),
|
125 |
+
"masked_sentence": datasets.Value("string"),
|
126 |
+
}
|
127 |
+
)
|
128 |
+
return datasets.DatasetInfo(
|
129 |
+
description=_DESCRIPTION,
|
130 |
+
features=features,
|
131 |
+
supervised_keys=None,
|
132 |
+
homepage=_HOMEPAGE,
|
133 |
+
license=_LICENSE,
|
134 |
+
citation=_CITATION,
|
135 |
+
)
|
136 |
+
elif self.config.name == "google_re":
|
137 |
+
features = datasets.Features(
|
138 |
+
{
|
139 |
+
"pred": datasets.Value("string"),
|
140 |
+
"sub": datasets.Value("string"),
|
141 |
+
"obj": datasets.Value("string"),
|
142 |
+
"evidences": datasets.Value("string"),
|
143 |
+
"judgments": datasets.Value("string"),
|
144 |
+
"sub_w": datasets.Value("string"),
|
145 |
+
"sub_label": datasets.Value("string"),
|
146 |
+
"sub_aliases": datasets.Value("string"),
|
147 |
+
"obj_w": datasets.Value("string"),
|
148 |
+
"obj_label": datasets.Value("string"),
|
149 |
+
"obj_aliases": datasets.Value("string"),
|
150 |
+
"uuid": datasets.Value("string"),
|
151 |
+
"masked_sentence": datasets.Value("string"),
|
152 |
+
"template": datasets.Value("string"),
|
153 |
+
"template_negated": datasets.Value("string"),
|
154 |
+
}
|
155 |
+
)
|
156 |
+
return datasets.DatasetInfo(
|
157 |
+
description=_DESCRIPTION,
|
158 |
+
features=features,
|
159 |
+
supervised_keys=None,
|
160 |
+
homepage=_HOMEPAGE,
|
161 |
+
license=_LICENSE,
|
162 |
+
citation=_CITATION,
|
163 |
+
)
|
164 |
+
|
165 |
+
def _split_generators(self, dl_manager):
|
166 |
+
"""Returns SplitGenerators."""
|
167 |
+
archive = dl_manager.download(_DATA_URL)
|
168 |
+
if self.config.name == "trex":
|
169 |
+
relations_path = dl_manager.download(_RELATIONS_URL)
|
170 |
+
return [
|
171 |
+
datasets.SplitGenerator(
|
172 |
+
name=datasets.Split.TRAIN,
|
173 |
+
gen_kwargs={
|
174 |
+
"filepaths": ["TREx/*"],
|
175 |
+
"files": dl_manager.iter_archive(archive),
|
176 |
+
"relations_path": relations_path,
|
177 |
+
},
|
178 |
+
),
|
179 |
+
]
|
180 |
+
elif self.config.name == "google_re":
|
181 |
+
return [
|
182 |
+
datasets.SplitGenerator(
|
183 |
+
name=datasets.Split.TRAIN,
|
184 |
+
gen_kwargs={
|
185 |
+
"filepaths": [
|
186 |
+
"Google_RE/date_of_birth_test.jsonl",
|
187 |
+
"Google_RE/place_of_birth_test.jsonl",
|
188 |
+
"Google_RE/place_of_death_test.jsonl",
|
189 |
+
],
|
190 |
+
"files": dl_manager.iter_archive(archive),
|
191 |
+
},
|
192 |
+
),
|
193 |
+
]
|
194 |
+
elif self.config.name == "conceptnet":
|
195 |
+
return [
|
196 |
+
datasets.SplitGenerator(
|
197 |
+
name=datasets.Split.TRAIN,
|
198 |
+
gen_kwargs={
|
199 |
+
"filepaths": ["ConceptNet/test.jsonl"],
|
200 |
+
"files": dl_manager.iter_archive(archive),
|
201 |
+
},
|
202 |
+
),
|
203 |
+
]
|
204 |
+
elif self.config.name == "squad":
|
205 |
+
return [
|
206 |
+
datasets.SplitGenerator(
|
207 |
+
name=datasets.Split.TRAIN,
|
208 |
+
gen_kwargs={
|
209 |
+
"filepaths": ["Squad/test.jsonl"],
|
210 |
+
"files": dl_manager.iter_archive(archive),
|
211 |
+
},
|
212 |
+
),
|
213 |
+
]
|
214 |
+
|
215 |
+
def _generate_examples(self, filepaths, files, relations_path=None):
|
216 |
+
"""Yields examples from the LAMA dataset."""
|
217 |
+
filepaths = list(filepaths)
|
218 |
+
if self.config.name == "trex":
|
219 |
+
all_rels = {}
|
220 |
+
with open(relations_path, encoding="utf-8") as f:
|
221 |
+
for row in f:
|
222 |
+
data = json.loads(row)
|
223 |
+
all_rels[data["relation"]] = data
|
224 |
+
id_ = -1
|
225 |
+
inside_trec_directory = False
|
226 |
+
for path, f in files:
|
227 |
+
if any(fnmatch(path, pattern) for pattern in filepaths):
|
228 |
+
inside_trec_directory = True
|
229 |
+
for row in f:
|
230 |
+
data = json.loads(row)
|
231 |
+
pred = all_rels.get(data["predicate_id"], {})
|
232 |
+
for evidences in data["evidences"]:
|
233 |
+
id_ += 1
|
234 |
+
yield id_, {
|
235 |
+
"uuid": str(data["uuid"]),
|
236 |
+
"obj_uri": str(data["obj_uri"]),
|
237 |
+
"obj_label": str(data["obj_label"]),
|
238 |
+
"sub_uri": str(data["sub_uri"]),
|
239 |
+
"sub_label": str(data["sub_label"]),
|
240 |
+
"predicate_id": str(data["predicate_id"]),
|
241 |
+
"sub_surface": str(evidences["sub_surface"]),
|
242 |
+
"obj_surface": str(evidences["obj_surface"]),
|
243 |
+
"masked_sentence": str(evidences["masked_sentence"]),
|
244 |
+
"template": str(pred.get("template", "")),
|
245 |
+
"template_negated": str(pred.get("template_negated", "")),
|
246 |
+
"label": str(pred.get("label", "")),
|
247 |
+
"description": str(pred.get("description", "")),
|
248 |
+
"type": str(pred.get("type", "")),
|
249 |
+
}
|
250 |
+
elif inside_trec_directory:
|
251 |
+
break
|
252 |
+
elif self.config.name == "conceptnet":
|
253 |
+
id_ = -1
|
254 |
+
for path, f in files:
|
255 |
+
if not filepaths:
|
256 |
+
break
|
257 |
+
if path in list(filepaths):
|
258 |
+
for row in f:
|
259 |
+
data = json.loads(row)
|
260 |
+
if data.get("negated") is not None:
|
261 |
+
for masked_sentence, negated in zip(data["masked_sentences"], data["negated"]):
|
262 |
+
id_ += 1
|
263 |
+
yield id_, {
|
264 |
+
"uuid": str(data["uuid"]),
|
265 |
+
"sub": str(data.get("sub", "")),
|
266 |
+
"obj": str(data.get("obj", "")),
|
267 |
+
"pred": str(data["pred"]),
|
268 |
+
"obj_label": str(data["obj_label"]),
|
269 |
+
"masked_sentence": str(masked_sentence),
|
270 |
+
"negated": str(negated),
|
271 |
+
}
|
272 |
+
else:
|
273 |
+
for masked_sentence in data["masked_sentences"]:
|
274 |
+
id_ += 1
|
275 |
+
yield id_, {
|
276 |
+
"uuid": str(data["uuid"]),
|
277 |
+
"sub": str(data.get("sub", "")),
|
278 |
+
"obj": str(data.get("obj", "")),
|
279 |
+
"pred": str(data["pred"]),
|
280 |
+
"obj_label": str(data["obj_label"]),
|
281 |
+
"masked_sentence": str(masked_sentence),
|
282 |
+
"negated": str(""),
|
283 |
+
}
|
284 |
+
filepaths.remove(path)
|
285 |
+
elif self.config.name == "squad":
|
286 |
+
id_ = -1
|
287 |
+
for path, f in files:
|
288 |
+
if not filepaths:
|
289 |
+
break
|
290 |
+
if path in filepaths:
|
291 |
+
for row in f:
|
292 |
+
data = json.loads(row)
|
293 |
+
for masked_sentence in data["masked_sentences"]:
|
294 |
+
id_ += 1
|
295 |
+
yield id_, {
|
296 |
+
"id": str(data["id"]),
|
297 |
+
"sub_label": str(data["sub_label"]),
|
298 |
+
"obj_label": str(data["obj_label"]),
|
299 |
+
"negated": str(data.get("negated", "")),
|
300 |
+
"masked_sentence": str(masked_sentence),
|
301 |
+
}
|
302 |
+
filepaths.remove(path)
|
303 |
+
elif self.config.name == "google_re":
|
304 |
+
id_ = -1
|
305 |
+
for path, f in files:
|
306 |
+
if path in filepaths:
|
307 |
+
if not filepaths:
|
308 |
+
break
|
309 |
+
if path in filepaths:
|
310 |
+
# from https://github.com/facebookresearch/LAMA/blob/master/scripts/run_experiments.py
|
311 |
+
if "place_of_birth" in path:
|
312 |
+
pred = {
|
313 |
+
"relation": "place_of_birth",
|
314 |
+
"template": "[X] was born in [Y] .",
|
315 |
+
"template_negated": "[X] was not born in [Y] .",
|
316 |
+
}
|
317 |
+
elif "date_of_birth" in path:
|
318 |
+
pred = {
|
319 |
+
"relation": "date_of_birth",
|
320 |
+
"template": "[X] (born [Y]).",
|
321 |
+
"template_negated": "[X] (not born [Y]).",
|
322 |
+
}
|
323 |
+
else:
|
324 |
+
pred = {
|
325 |
+
"relation": "place_of_death",
|
326 |
+
"template": "[X] died in [Y] .",
|
327 |
+
"template_negated": "[X] did not die in [Y] .",
|
328 |
+
}
|
329 |
+
for row in f:
|
330 |
+
data = json.loads(row)
|
331 |
+
for masked_sentence in data["masked_sentences"]:
|
332 |
+
id_ += 1
|
333 |
+
yield id_, {
|
334 |
+
"pred": str(data["pred"]),
|
335 |
+
"sub": str(data["sub"]),
|
336 |
+
"obj": str(data["obj"]),
|
337 |
+
"evidences": str(data["evidences"]),
|
338 |
+
"judgments": str(data["judgments"]),
|
339 |
+
"sub_w": str(data["sub_w"]),
|
340 |
+
"sub_label": str(data["sub_label"]),
|
341 |
+
"sub_aliases": str(data["sub_aliases"]),
|
342 |
+
"obj_w": str(data["obj_w"]),
|
343 |
+
"obj_label": str(data["obj_label"]),
|
344 |
+
"obj_aliases": str(data["obj_aliases"]),
|
345 |
+
"uuid": str(data["uuid"]),
|
346 |
+
"masked_sentence": str(masked_sentence),
|
347 |
+
"template": str(pred["template"]),
|
348 |
+
"template_negated": str(pred["template_negated"]),
|
349 |
+
}
|
350 |
+
filepaths.remove(path)
|