Datasets:

Modalities:
Text
Languages:
English
Size:
< 1K
Libraries:
Datasets
License:
gabrielaltay commited on
Commit
24c19f4
1 Parent(s): 9367da5

upload hubscripts/lll_hub.py to hub from bigbio repo

Browse files
Files changed (1) hide show
  1. lll.py +328 -0
lll.py ADDED
@@ -0,0 +1,328 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2022 The HuggingFace Datasets Authors and Simon Ott, github: nomisto
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
+
16
+ """
17
+ The LLL05 challenge task is to learn rules to extract protein/gene interactions from biology abstracts from the Medline
18
+ bibliography database. The goal of the challenge is to test the ability of the participating IE systems to identify the
19
+ interactions and the gene/proteins that interact. The participants will test their IE patterns on a test set with the
20
+ aim of extracting the correct agent and target.The challenge focuses on information extraction of gene interactions in
21
+ Bacillus subtilis. Extracting gene interaction is the most popular event IE task in biology. Bacillus subtilis (Bs) is
22
+ a model bacterium and many papers have been published on direct gene interactions involved in sporulation. The gene
23
+ interactions are generally mentioned in the abstract and the full text of the paper is not needed. Extracting gene
24
+ interaction means, extracting the agent (proteins) and the target (genes) of all couples of genic interactions from
25
+ sentences.
26
+ """
27
+
28
+ # NOTE:
29
+ # word stop offsets are increased by one to be consistent with python slicing.
30
+ # test set does not include entity relation information
31
+
32
+ import itertools as it
33
+ from typing import List
34
+
35
+ import datasets
36
+
37
+ from .bigbiohub import kb_features
38
+ from .bigbiohub import BigBioConfig
39
+ from .bigbiohub import Tasks
40
+
41
+ _LANGUAGES = ['English']
42
+ _PUBMED = True
43
+ _LOCAL = False
44
+ _CITATION = """\
45
+ @article{article,
46
+ author = {Nédellec, C.},
47
+ year = {2005},
48
+ month = {01},
49
+ pages = {},
50
+ title = {Learning Language in Logic - Genic Interaction Extraction Challenge},
51
+ journal = {Proceedings of the Learning Language in Logic 2005 Workshop at the \
52
+ International Conference on Machine Learning}
53
+ }
54
+ """
55
+
56
+ _DATASETNAME = "lll"
57
+ _DISPLAYNAME = "LLL05"
58
+
59
+ _DESCRIPTION = """\
60
+ The LLL05 challenge task is to learn rules to extract protein/gene interactions from biology abstracts from the Medline
61
+ bibliography database. The goal of the challenge is to test the ability of the participating IE systems to identify the
62
+ interactions and the gene/proteins that interact. The participants will test their IE patterns on a test set with the
63
+ aim of extracting the correct agent and target.The challenge focuses on information extraction of gene interactions in
64
+ Bacillus subtilis. Extracting gene interaction is the most popular event IE task in biology. Bacillus subtilis (Bs) is
65
+ a model bacterium and many papers have been published on direct gene interactions involved in sporulation. The gene
66
+ interactions are generally mentioned in the abstract and the full text of the paper is not needed. Extracting gene
67
+ interaction means, extracting the agent (proteins) and the target (genes) of all couples of genic interactions from
68
+ sentences.
69
+ """
70
+
71
+ _HOMEPAGE = "http://genome.jouy.inra.fr/texte/LLLchallenge"
72
+
73
+ _LICENSE = 'License information unavailable'
74
+
75
+ _URLS = {
76
+ _DATASETNAME: [
77
+ "http://genome.jouy.inra.fr/texte/LLLchallenge/data/LLLChalenge05/data/train/task2/genic_interaction_linguistic_data.txt", # noqa
78
+ "http://genome.jouy.inra.fr/texte/LLLchallenge/data/LLLChalenge05/data/train/task2/genic_interaction_linguistic_data_coref.txt", # noqa
79
+ "http://genome.jouy.inra.fr/texte/LLLchallenge/data/LLLChalenge05/data/test/task2/enriched_test_data.txt", # noqa
80
+ ]
81
+ }
82
+
83
+ _SUPPORTED_TASKS = [Tasks.RELATION_EXTRACTION]
84
+
85
+ _SOURCE_VERSION = "1.0.0"
86
+
87
+ _BIGBIO_VERSION = "1.0.0"
88
+
89
+
90
+ class LLLDataset(datasets.GeneratorBasedBuilder):
91
+ """LLL dataset for gene interaction extraction (RE)"""
92
+
93
+ SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
94
+ BIGBIO_VERSION = datasets.Version(_BIGBIO_VERSION)
95
+
96
+ BUILDER_CONFIGS = [
97
+ BigBioConfig(
98
+ name="lll_source",
99
+ version=SOURCE_VERSION,
100
+ description="LLL source schema",
101
+ schema="source",
102
+ subset_id="lll",
103
+ ),
104
+ BigBioConfig(
105
+ name="lll_bigbio_kb",
106
+ version=BIGBIO_VERSION,
107
+ description="LLL BigBio schema",
108
+ schema="bigbio_kb",
109
+ subset_id="lll",
110
+ ),
111
+ ]
112
+
113
+ DEFAULT_CONFIG_NAME = "lll_source"
114
+
115
+ def _info(self) -> datasets.DatasetInfo:
116
+
117
+ if self.config.schema == "source":
118
+ features = datasets.Features(
119
+ {
120
+ "id": datasets.Value("string"),
121
+ "sentence": datasets.Value("string"),
122
+ "words": [
123
+ {
124
+ "id": datasets.Value("string"),
125
+ "text": datasets.Value("string"),
126
+ "offsets": datasets.Sequence(datasets.Value("int32")),
127
+ }
128
+ ],
129
+ "genic_interactions": [
130
+ {
131
+ "ref_id1": datasets.Value("string"),
132
+ "ref_id2": datasets.Value("string"),
133
+ }
134
+ ],
135
+ "agents": [
136
+ {
137
+ "ref_id": datasets.Value("string"),
138
+ }
139
+ ],
140
+ "targets": [
141
+ {
142
+ "ref_id": datasets.Value("string"),
143
+ }
144
+ ],
145
+ "lemmas": [
146
+ {
147
+ "ref_id": datasets.Value("string"),
148
+ "lemma": datasets.Value("string"),
149
+ }
150
+ ],
151
+ "syntactic_relations": [
152
+ {
153
+ "type": datasets.Value("string"),
154
+ "ref_id1": datasets.Value("string"),
155
+ "ref_id2": datasets.Value("string"),
156
+ }
157
+ ],
158
+ }
159
+ )
160
+
161
+ elif self.config.schema == "bigbio_kb":
162
+ features = kb_features
163
+
164
+ return datasets.DatasetInfo(
165
+ description=_DESCRIPTION,
166
+ features=features,
167
+ homepage=_HOMEPAGE,
168
+ license=str(_LICENSE),
169
+ citation=_CITATION,
170
+ )
171
+
172
+ def _split_generators(self, dl_manager) -> List[datasets.SplitGenerator]:
173
+
174
+ urls = _URLS[_DATASETNAME]
175
+ train_path, train_coref_path, test_path = dl_manager.download_and_extract(urls)
176
+
177
+ return [
178
+ datasets.SplitGenerator(
179
+ name=datasets.Split.TRAIN,
180
+ gen_kwargs={
181
+ "data_paths": [train_path, train_coref_path],
182
+ "split": "train",
183
+ },
184
+ ),
185
+ datasets.SplitGenerator(
186
+ name=datasets.Split.TEST,
187
+ gen_kwargs={"data_paths": [test_path], "split": "test"},
188
+ ),
189
+ ]
190
+
191
+ def _generate_examples(self, data_paths, split):
192
+
193
+ if self.config.schema == "source":
194
+ for path in data_paths:
195
+ with open(path, encoding="utf8") as documents:
196
+ for document in self._generate_parsed_documents(documents, split):
197
+ yield document["id"], document
198
+
199
+ elif self.config.schema == "bigbio_kb":
200
+ uid = it.count(0)
201
+ for path in data_paths:
202
+ with open(path, encoding="utf8") as documents:
203
+ for document in self._generate_parsed_documents(documents, split):
204
+ document_ = {}
205
+ document_["id"] = next(uid)
206
+ document_["document_id"] = document["id"]
207
+
208
+ document_["passages"] = [
209
+ {
210
+ "id": next(uid),
211
+ "type": BigBioValues.NULL,
212
+ "text": [document["sentence"]],
213
+ "offsets": [[0, len(document["sentence"])]],
214
+ }
215
+ ]
216
+
217
+ id_to_word = {i["id"]: i for i in document["words"]}
218
+ document_["entities"] = []
219
+ for agent in document["agents"]:
220
+ word = id_to_word[agent["ref_id"]]
221
+ document_["entities"].append(
222
+ {
223
+ "id": f"{document_['id']}-agent-{word['id']}",
224
+ "type": "agent",
225
+ "text": [word["text"]],
226
+ "offsets": [
227
+ [word["offsets"][0], word["offsets"][1]]
228
+ ],
229
+ "normalized": [],
230
+ }
231
+ )
232
+ for agent in document["targets"]:
233
+ word = id_to_word[agent["ref_id"]]
234
+ document_["entities"].append(
235
+ {
236
+ "id": f"{document_['id']}-target-{word['id']}",
237
+ "type": "target",
238
+ "text": [word["text"]],
239
+ "offsets": [
240
+ [word["offsets"][0], word["offsets"][1]]
241
+ ],
242
+ "normalized": [],
243
+ }
244
+ )
245
+
246
+ document_["relations"] = [
247
+ {
248
+ "id": next(uid),
249
+ "type": "genic_interaction",
250
+ "arg1_id": f"{document_['id']}-agent-{relation['ref_id1']}",
251
+ "arg2_id": f"{document_['id']}-target-{relation['ref_id2']}",
252
+ "normalized": [],
253
+ }
254
+ for relation in document["genic_interactions"]
255
+ ]
256
+
257
+ document_["events"] = []
258
+ document_["coreferences"] = []
259
+ yield document_["document_id"], document_
260
+
261
+ def _generate_parsed_documents(self, fstream, split):
262
+ for raw_document in self._generate_raw_documents(fstream):
263
+ yield self._parse_document(raw_document, split)
264
+
265
+ def _generate_raw_documents(self, fstream):
266
+ raw_document = []
267
+ for line in fstream:
268
+ if "%" in line:
269
+ continue
270
+ elif line.strip():
271
+ raw_document.append(line.strip())
272
+ elif raw_document:
273
+ if raw_document:
274
+ yield raw_document
275
+ raw_document = []
276
+ # needed for last document
277
+ if raw_document:
278
+ yield raw_document
279
+
280
+ def _parse_document(self, raw_document, split):
281
+ document = {}
282
+ for line in raw_document:
283
+ key, value = line.split("\t", 1)
284
+ if key in ["ID", "sentence"]:
285
+ document[key.lower()] = value
286
+ elif key in [
287
+ "words",
288
+ "genic_interactions",
289
+ "agents",
290
+ "targets",
291
+ "lemmas",
292
+ "syntactic_relations",
293
+ ]:
294
+ document[key.lower()] = self._parse_elements(value, key)
295
+ else:
296
+ raise NotImplementedError()
297
+
298
+ # Needed as testset does not contain agents, targets and genic_interactions (dataset was part of a challenge)
299
+ if split == "test":
300
+ document.setdefault("genic_interactions", [])
301
+ document.setdefault("agents", [])
302
+ document.setdefault("targets", [])
303
+
304
+ return document
305
+
306
+ def _parse_elements(self, values, type):
307
+ return [self._parse_element(atom, type) for atom in values.split("\t")]
308
+
309
+ def _parse_element(self, atom, type):
310
+ # Sorry for that abomination, parses the arguments from atoms like rel(arg1, ..., argn)
311
+ args = atom.split("(", 1)[1][:-1].split(",")
312
+ if type == "words":
313
+ # fix offsets for python slicing
314
+ return {
315
+ "id": args[0],
316
+ "text": args[1].strip("'"),
317
+ "offsets": [int(args[2]), int(args[3]) + 1],
318
+ }
319
+ elif type == "genic_interactions":
320
+ return {"ref_id1": args[0], "ref_id2": args[1]}
321
+ elif type == "agents":
322
+ return {"ref_id": args[0]}
323
+ elif type == "targets":
324
+ return {"ref_id": args[0]}
325
+ elif type == "lemmas":
326
+ return {"ref_id": args[0], "lemma": args[1].strip("'")}
327
+ elif type == "syntactic_relations":
328
+ return {"type": args[0].strip("'"), "ref_id1": args[1], "ref_id2": args[2]}