Commit
·
1660668
1
Parent(s):
2606565
upload hubscripts/mlee_hub.py to hub from bigbio repo
Browse files
mlee.py
ADDED
@@ -0,0 +1,279 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
|
16 |
+
"""
|
17 |
+
MLEE is an event extraction corpus consisting of manually annotated abstracts of papers
|
18 |
+
on angiogenesis. It contains annotations for entities, relations, events and coreferences
|
19 |
+
The annotations span molecular, cellular, tissue, and organ-level processes.
|
20 |
+
"""
|
21 |
+
from pathlib import Path
|
22 |
+
from typing import Dict, List
|
23 |
+
|
24 |
+
import datasets
|
25 |
+
|
26 |
+
from .bigbiohub import kb_features
|
27 |
+
from .bigbiohub import BigBioConfig
|
28 |
+
from .bigbiohub import Tasks
|
29 |
+
|
30 |
+
_SOURCE_VIEW_NAME = "source"
|
31 |
+
_UNIFIED_VIEW_NAME = "bigbio"
|
32 |
+
|
33 |
+
_LANGUAGES = ['English']
|
34 |
+
_PUBMED = True
|
35 |
+
_LOCAL = False
|
36 |
+
_CITATION = """\
|
37 |
+
@article{pyysalo2012event,
|
38 |
+
title={Event extraction across multiple levels of biological organization},
|
39 |
+
author={Pyysalo, Sampo and Ohta, Tomoko and Miwa, Makoto and Cho, Han-Cheol and Tsujii, Jun'ichi and Ananiadou, Sophia},
|
40 |
+
journal={Bioinformatics},
|
41 |
+
volume={28},
|
42 |
+
number={18},
|
43 |
+
pages={i575--i581},
|
44 |
+
year={2012},
|
45 |
+
publisher={Oxford University Press}
|
46 |
+
}
|
47 |
+
"""
|
48 |
+
|
49 |
+
_DESCRIPTION = """\
|
50 |
+
MLEE is an event extraction corpus consisting of manually annotated abstracts of papers
|
51 |
+
on angiogenesis. It contains annotations for entities, relations, events and coreferences
|
52 |
+
The annotations span molecular, cellular, tissue, and organ-level processes.
|
53 |
+
"""
|
54 |
+
|
55 |
+
_DATASETNAME = "mlee"
|
56 |
+
_DISPLAYNAME = "MLEE"
|
57 |
+
|
58 |
+
_HOMEPAGE = "http://www.nactem.ac.uk/MLEE/"
|
59 |
+
|
60 |
+
_LICENSE = 'Creative Commons Attribution Non Commercial Share Alike 3.0 Unported'
|
61 |
+
_URLs = {
|
62 |
+
"source": "http://www.nactem.ac.uk/MLEE/MLEE-1.0.2-rev1.tar.gz",
|
63 |
+
"bigbio_kb": "http://www.nactem.ac.uk/MLEE/MLEE-1.0.2-rev1.tar.gz",
|
64 |
+
}
|
65 |
+
|
66 |
+
_SUPPORTED_TASKS = [
|
67 |
+
Tasks.EVENT_EXTRACTION,
|
68 |
+
Tasks.NAMED_ENTITY_RECOGNITION,
|
69 |
+
Tasks.RELATION_EXTRACTION,
|
70 |
+
Tasks.COREFERENCE_RESOLUTION,
|
71 |
+
]
|
72 |
+
_SOURCE_VERSION = "1.0.0"
|
73 |
+
_BIGBIO_VERSION = "1.0.0"
|
74 |
+
|
75 |
+
|
76 |
+
class MLEE(datasets.GeneratorBasedBuilder):
|
77 |
+
"""Write a short docstring documenting what this dataset is"""
|
78 |
+
|
79 |
+
SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
|
80 |
+
BIGBIO_VERSION = datasets.Version(_BIGBIO_VERSION)
|
81 |
+
|
82 |
+
BUILDER_CONFIGS = [
|
83 |
+
BigBioConfig(
|
84 |
+
name="mlee_source",
|
85 |
+
version=SOURCE_VERSION,
|
86 |
+
description="MLEE source schema",
|
87 |
+
schema="source",
|
88 |
+
subset_id="mlee",
|
89 |
+
),
|
90 |
+
BigBioConfig(
|
91 |
+
name="mlee_bigbio_kb",
|
92 |
+
version=SOURCE_VERSION,
|
93 |
+
description="MLEE BigBio schema",
|
94 |
+
schema="bigbio_kb",
|
95 |
+
subset_id="mlee",
|
96 |
+
),
|
97 |
+
]
|
98 |
+
|
99 |
+
DEFAULT_CONFIG_NAME = "mlee_source"
|
100 |
+
|
101 |
+
_ROLE_MAPPING = {
|
102 |
+
"Theme2": "Theme",
|
103 |
+
"Instrument2": "Instrument",
|
104 |
+
"Participant2": "Participant",
|
105 |
+
"Participant3": "Participant",
|
106 |
+
"Participant4": "Participant",
|
107 |
+
}
|
108 |
+
|
109 |
+
def _info(self):
|
110 |
+
"""
|
111 |
+
Provide information about MLEE:
|
112 |
+
- `features` defines the schema of the parsed data set. The schema depends on the
|
113 |
+
chosen `config`: If it is `_SOURCE_VIEW_NAME` the schema is the schema of the
|
114 |
+
original data. If `config` is `_UNIFIED_VIEW_NAME`, then the schema is the
|
115 |
+
canonical KB-task schema defined in `biomedical/schemas/kb.py`.
|
116 |
+
|
117 |
+
"""
|
118 |
+
if self.config.schema == "source":
|
119 |
+
features = datasets.Features(
|
120 |
+
{
|
121 |
+
"id": datasets.Value("string"),
|
122 |
+
"document_id": datasets.Value("string"),
|
123 |
+
"text": datasets.Value("string"),
|
124 |
+
"text_bound_annotations": [ # T line in brat, e.g. type or event trigger
|
125 |
+
{
|
126 |
+
"offsets": datasets.Sequence([datasets.Value("int32")]),
|
127 |
+
"text": datasets.Sequence(datasets.Value("string")),
|
128 |
+
"type": datasets.Value("string"),
|
129 |
+
"id": datasets.Value("string"),
|
130 |
+
}
|
131 |
+
],
|
132 |
+
"events": [ # E line in brat
|
133 |
+
{
|
134 |
+
"trigger": datasets.Value(
|
135 |
+
"string"
|
136 |
+
), # refers to the text_bound_annotation of the trigger,
|
137 |
+
"id": datasets.Value("string"),
|
138 |
+
"type": datasets.Value("string"),
|
139 |
+
"arguments": datasets.Sequence(
|
140 |
+
{
|
141 |
+
"role": datasets.Value("string"),
|
142 |
+
"ref_id": datasets.Value("string"),
|
143 |
+
}
|
144 |
+
),
|
145 |
+
}
|
146 |
+
],
|
147 |
+
"relations": [ # R line in brat
|
148 |
+
{
|
149 |
+
"id": datasets.Value("string"),
|
150 |
+
"head": {
|
151 |
+
"ref_id": datasets.Value("string"),
|
152 |
+
"role": datasets.Value("string"),
|
153 |
+
},
|
154 |
+
"tail": {
|
155 |
+
"ref_id": datasets.Value("string"),
|
156 |
+
"role": datasets.Value("string"),
|
157 |
+
},
|
158 |
+
"type": datasets.Value("string"),
|
159 |
+
}
|
160 |
+
],
|
161 |
+
"equivalences": [ # Equiv line in brat
|
162 |
+
{
|
163 |
+
"id": datasets.Value("string"),
|
164 |
+
"ref_ids": datasets.Sequence(datasets.Value("string")),
|
165 |
+
}
|
166 |
+
],
|
167 |
+
"attributes": [ # M or A lines in brat
|
168 |
+
{
|
169 |
+
"id": datasets.Value("string"),
|
170 |
+
"type": datasets.Value("string"),
|
171 |
+
"ref_id": datasets.Value("string"),
|
172 |
+
"value": datasets.Value("string"),
|
173 |
+
}
|
174 |
+
],
|
175 |
+
"normalizations": [ # N lines in brat
|
176 |
+
{
|
177 |
+
"id": datasets.Value("string"),
|
178 |
+
"type": datasets.Value("string"),
|
179 |
+
"ref_id": datasets.Value("string"),
|
180 |
+
"resource_name": datasets.Value(
|
181 |
+
"string"
|
182 |
+
), # Name of the resource, e.g. "Wikipedia"
|
183 |
+
"cuid": datasets.Value(
|
184 |
+
"string"
|
185 |
+
), # ID in the resource, e.g. 534366
|
186 |
+
"text": datasets.Value(
|
187 |
+
"string"
|
188 |
+
), # Human readable description/name of the entity, e.g. "Barack Obama"
|
189 |
+
}
|
190 |
+
],
|
191 |
+
},
|
192 |
+
)
|
193 |
+
elif self.config.schema == "bigbio_kb":
|
194 |
+
features = kb_features
|
195 |
+
|
196 |
+
return datasets.DatasetInfo(
|
197 |
+
# This is the description that will appear on the datasets page.
|
198 |
+
description=_DESCRIPTION,
|
199 |
+
features=features,
|
200 |
+
# If there's a common (input, target) tuple from the features, uncomment supervised_keys line below and
|
201 |
+
# specify them. They'll be used if as_supervised=True in builder.as_dataset.
|
202 |
+
# This is not applicable for MLEE.
|
203 |
+
# supervised_keys=("sentence", "label"),
|
204 |
+
# Homepage of the dataset for documentation
|
205 |
+
homepage=_HOMEPAGE,
|
206 |
+
# License for the dataset if available
|
207 |
+
license=str(_LICENSE),
|
208 |
+
# Citation for the dataset
|
209 |
+
citation=_CITATION,
|
210 |
+
)
|
211 |
+
|
212 |
+
def _split_generators(
|
213 |
+
self, dl_manager: datasets.DownloadManager
|
214 |
+
) -> List[datasets.SplitGenerator]:
|
215 |
+
"""
|
216 |
+
Create the three splits provided by MLEE: train, validation and test.
|
217 |
+
|
218 |
+
Each split is created by instantiating a `datasets.SplitGenerator`, which will
|
219 |
+
call `this._generate_examples` with the keyword arguments in `gen_kwargs`.
|
220 |
+
"""
|
221 |
+
|
222 |
+
my_urls = _URLs[self.config.schema]
|
223 |
+
data_dir = Path(dl_manager.download_and_extract(my_urls))
|
224 |
+
data_files = {
|
225 |
+
"train": data_dir
|
226 |
+
/ "MLEE-1.0.2-rev1"
|
227 |
+
/ "standoff"
|
228 |
+
/ "development"
|
229 |
+
/ "train",
|
230 |
+
"dev": data_dir / "MLEE-1.0.2-rev1" / "standoff" / "development" / "test",
|
231 |
+
"test": data_dir / "MLEE-1.0.2-rev1" / "standoff" / "test" / "test",
|
232 |
+
}
|
233 |
+
|
234 |
+
return [
|
235 |
+
datasets.SplitGenerator(
|
236 |
+
name=datasets.Split.TRAIN,
|
237 |
+
gen_kwargs={"data_files": data_files["train"]},
|
238 |
+
),
|
239 |
+
datasets.SplitGenerator(
|
240 |
+
name=datasets.Split.VALIDATION,
|
241 |
+
gen_kwargs={"data_files": data_files["dev"]},
|
242 |
+
),
|
243 |
+
datasets.SplitGenerator(
|
244 |
+
name=datasets.Split.TEST,
|
245 |
+
gen_kwargs={"data_files": data_files["test"]},
|
246 |
+
),
|
247 |
+
]
|
248 |
+
|
249 |
+
def _standardize_arguments_roles(self, kb_example: Dict) -> Dict:
|
250 |
+
|
251 |
+
for event in kb_example["events"]:
|
252 |
+
for argument in event["arguments"]:
|
253 |
+
role = argument["role"]
|
254 |
+
argument["role"] = self._ROLE_MAPPING.get(role, role)
|
255 |
+
|
256 |
+
return kb_example
|
257 |
+
|
258 |
+
def _generate_examples(self, data_files: Path):
|
259 |
+
"""
|
260 |
+
Yield one `(guid, example)` pair per abstract in MLEE.
|
261 |
+
The contents of `example` will depend on the chosen configuration.
|
262 |
+
"""
|
263 |
+
if self.config.schema == "source":
|
264 |
+
txt_files = list(data_files.glob("*txt"))
|
265 |
+
for guid, txt_file in enumerate(txt_files):
|
266 |
+
example = parsing.parse_brat_file(txt_file)
|
267 |
+
example["id"] = str(guid)
|
268 |
+
yield guid, example
|
269 |
+
elif self.config.schema == "bigbio_kb":
|
270 |
+
txt_files = list(data_files.glob("*txt"))
|
271 |
+
for guid, txt_file in enumerate(txt_files):
|
272 |
+
example = parsing.brat_parse_to_bigbio_kb(
|
273 |
+
parsing.parse_brat_file(txt_file)
|
274 |
+
)
|
275 |
+
example = self._standardize_arguments_roles(example)
|
276 |
+
example["id"] = str(guid)
|
277 |
+
yield guid, example
|
278 |
+
else:
|
279 |
+
raise ValueError(f"Invalid config: {self.config.name}")
|