gabrielaltay commited on
Commit
8c0e245
1 Parent(s): d7eacb2

upload hub_repos/cardiode/cardiode.py to hub from bigbio repo

Browse files
Files changed (1) hide show
  1. cardiode.py +300 -0
cardiode.py ADDED
@@ -0,0 +1,300 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2022 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
+ import os
17
+ from pathlib import Path
18
+ from typing import Dict, List, Tuple
19
+
20
+ import datasets
21
+ import pandas as pd
22
+
23
+ from .bigbiohub import BigBioConfig, Tasks, kb_features
24
+
25
+ _LOCAL = True
26
+ _CITATION = """\
27
+ @data{
28
+ data/AFYQDY_2022,
29
+ author = {Christoph Dieterich},
30
+ publisher = {heiDATA},
31
+ title = {{CARDIO:DE}},
32
+ year = {2022},
33
+ version = {V5},
34
+ doi = {10.11588/data/AFYQDY},
35
+ url = {https://doi.org/10.11588/data/AFYQDY}
36
+ }
37
+ """
38
+
39
+ _DESCRIPTION = """\
40
+ First freely available and distributable large German clinical corpus from the cardiovascular domain.
41
+ """
42
+
43
+ _HOMEPAGE = "https://heidata.uni-heidelberg.de/dataset.xhtml?persistentId=doi%3A10.11588%2Fdata%2FAFYQDY"
44
+
45
+ _LICENSE = "DUA"
46
+ _LANGUAGES = ["German"]
47
+ _URLS = {}
48
+ _SUPPORTED_TASKS = [Tasks.NAMED_ENTITY_RECOGNITION]
49
+ _SOURCE_VERSION = "5.0.0"
50
+ _BIGBIO_VERSION = "1.0.0"
51
+ _DATASETNAME = "cardiode"
52
+ _DISPLAYNAME = "CARDIO:DE"
53
+ _PUBMED = False
54
+
55
+
56
+ class CardioDataset(datasets.GeneratorBasedBuilder):
57
+ SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
58
+ BIGBIO_VERSION = datasets.Version(_BIGBIO_VERSION)
59
+
60
+ BUILDER_CONFIGS = [
61
+ BigBioConfig(
62
+ name="cardiode_source",
63
+ version=SOURCE_VERSION,
64
+ description="CARDIO:DE source schema",
65
+ schema="source",
66
+ subset_id="cardiode",
67
+ ),
68
+ BigBioConfig(
69
+ name="cardiode_bigbio_kb",
70
+ version=BIGBIO_VERSION,
71
+ description="CARDIO:DE BigBio schema",
72
+ schema="bigbio_kb",
73
+ subset_id="cardidoe",
74
+ ),
75
+ ]
76
+
77
+ DEFAULT_CONFIG_NAME = "cardiode_bigbio_kb"
78
+
79
+ def _info(self) -> datasets.DatasetInfo:
80
+ if self.config.schema == "source":
81
+ features = datasets.Features(
82
+ {
83
+ "doc_id": datasets.Value("string"),
84
+ "annotations": [
85
+ {
86
+ "text": datasets.Value("string"),
87
+ "tokens": [
88
+ {
89
+ "id": datasets.Value("string"),
90
+ "offsets": datasets.Value("string"),
91
+ "text": datasets.Value("string"),
92
+ "type": datasets.Value("string"),
93
+ "parent_annotation_id": datasets.Value("string"),
94
+ "section": datasets.Value("string"),
95
+ }
96
+ ],
97
+ }
98
+ ],
99
+ }
100
+ )
101
+
102
+ elif self.config.schema == "bigbio_kb":
103
+ features = kb_features
104
+
105
+ return datasets.DatasetInfo(
106
+ description=_DESCRIPTION,
107
+ features=features,
108
+ homepage=_HOMEPAGE,
109
+ license=_LICENSE,
110
+ citation=_CITATION,
111
+ )
112
+
113
+ def _split_generators(self, dl_manager) -> List[datasets.SplitGenerator]:
114
+ if self.config.data_dir is None:
115
+ raise ValueError("This is a local dataset. Please pass the data_dir kwarg to load_dataset.")
116
+ else:
117
+ data_dir = self.config.data_dir
118
+
119
+ return [
120
+ datasets.SplitGenerator(
121
+ name=datasets.Split.TRAIN,
122
+ gen_kwargs={
123
+ "filepath": os.path.join(data_dir),
124
+ "split": "train",
125
+ },
126
+ )
127
+ ]
128
+
129
+ def _generate_examples(self, filepath, split: str) -> Tuple[int, Dict]:
130
+ """Yields examples as (key, example) tuples."""
131
+ doc_ids = _sort_files(Path(filepath) / "tsv" / "CARDIODE400_main")
132
+ for uid, doc in enumerate(doc_ids):
133
+ tsv_path = Path(filepath) / "tsv" / "CARDIODE400_main" / f"{doc}"
134
+ df, sentences = _parse_tsv(tsv_path)
135
+ if self.config.schema == "source":
136
+ yield uid, _make_source(uid, doc, df, sentences)
137
+ elif self.config.schema == "bigbio_kb":
138
+ yield uid, _make_bigbio_kb(uid, doc, df, sentences)
139
+
140
+
141
+ def _parse_tsv(path: str) -> pd.DataFrame:
142
+ # read whole .tsv as a string
143
+ with open(path, encoding="utf-8") as file:
144
+ content = file.read()
145
+
146
+ # separate doc into sentences
147
+ passages = content.split("#")
148
+
149
+ # remove the first line (un-tabbed) of each sentence
150
+ # split sentences into words/tokens
151
+ # and store string sentences for the passages
152
+ sentences = []
153
+ for i, passage in enumerate(passages):
154
+ if passage.split("\n")[0].startswith("Text="):
155
+ sentences.append(passage.split("\n")[0].split("Text=")[1])
156
+ passages[i] = passage.split("\n")[1:]
157
+
158
+ # clean empty sentences and tokens
159
+ clean_passages = [[token for token in passage if token != ""] for passage in passages if passage != []]
160
+
161
+ # make a dataframe out of the clean tokens
162
+ df = []
163
+ for passage in clean_passages:
164
+ for token in passage:
165
+ df.append(token.split("\t"))
166
+
167
+ df = pd.DataFrame(df).rename(
168
+ columns={
169
+ 0: "passage_token_id",
170
+ 1: "token_offset",
171
+ 2: "text",
172
+ 3: "label",
173
+ 4: "uncertain",
174
+ 5: "relation",
175
+ 6: "section",
176
+ }
177
+ )
178
+
179
+ # correct weird rows were label is NoneType
180
+ df["label"].fillna("_", inplace=True)
181
+
182
+ # split passage and token ids
183
+ df[["passage_id", "token_id"]] = df["passage_token_id"].str.split("-", expand=True)
184
+
185
+ # split labels and their spans
186
+ # some docs do not have labels spanning various tokens (or they do not have any labels at all)
187
+ if df["label"].apply(lambda x: "[" in x).any():
188
+ df[["lab", "span"]] = df["label"].str.split("[", expand=True)
189
+ df["span"] = df["span"].str.replace("]", "", regex=True)
190
+ else:
191
+ df["lab"] = "_"
192
+ df["span"] = None
193
+
194
+ # split start and end offsets and cast to int
195
+ df[["offset_start", "offset_end"]] = df["token_offset"].str.split("-", expand=True)
196
+ df["offset_start"] = df["offset_start"].astype(int)
197
+ df["offset_end"] = df["offset_end"].astype(int)
198
+
199
+ # correct offset gaps between tokens
200
+ i = 0
201
+ while i < len(df) - 1:
202
+ gap = df.loc[i + 1]["offset_start"] - df.loc[i]["offset_end"]
203
+ if gap > 1:
204
+ df.loc[i + 1 :, "offset_start"] = df.loc[i + 1 :, "offset_start"] - (gap - 1)
205
+ df.loc[i + 1 :, "offset_end"] = df.loc[i + 1 :, "offset_end"] - (gap - 1)
206
+ i += 1
207
+
208
+ return df, sentences
209
+
210
+
211
+ def _make_source(uid: int, doc_id: str, df: pd.DataFrame, sentences: list):
212
+ out = {"doc_id": doc_id, "annotations": []}
213
+ for i, sentence in enumerate(sentences):
214
+ anno = {"text": sentence, "tokens": []}
215
+ chunk = df[df["passage_id"] == str(i + 1)]
216
+ for _, row in chunk.iterrows():
217
+ anno["tokens"].append(
218
+ {
219
+ "id": row["passage_token_id"],
220
+ "offsets": row["token_offset"],
221
+ "text": row["text"],
222
+ "type": row["label"],
223
+ "parent_annotation_id": row["relation"],
224
+ "section": row["section"],
225
+ }
226
+ )
227
+ out["annotations"].append(anno)
228
+ return out
229
+
230
+
231
+ def _make_bigbio_kb(uid: int, doc_id: str, df: pd.DataFrame, sentences: list):
232
+ out = {
233
+ "id": str(uid),
234
+ "document_id": doc_id,
235
+ "passages": [],
236
+ "entities": [],
237
+ "events": [],
238
+ "coreferences": [],
239
+ "relations": [],
240
+ }
241
+
242
+ # handle passages
243
+ i, sen_num, offset_mark = 0, 0, 0
244
+ while i < len(df):
245
+ pid = df.iloc[i]["passage_id"]
246
+ passage = df[df["passage_id"] == pid]
247
+
248
+ out["passages"].append(
249
+ {
250
+ "id": f"{uid}-{pid}",
251
+ "type": "sentence",
252
+ "text": [sentences[sen_num]],
253
+ "offsets": [[offset_mark, offset_mark + len(sentences[sen_num])]],
254
+ }
255
+ )
256
+
257
+ i += len(passage)
258
+ offset_mark += len(sentences[sen_num]) + 1
259
+ sen_num += 1
260
+
261
+ # handle entities
262
+ text = " ".join(sentences)
263
+ i = 0
264
+ while i < len(df):
265
+ if df.iloc[i]["lab"] != "_" and df.iloc[i]["span"] is None:
266
+ out["entities"].append(
267
+ {
268
+ "id": f'{uid}-{df.iloc[i]["passage_token_id"]}',
269
+ "type": df.iloc[i]["lab"],
270
+ "text": [text[df.iloc[i]["offset_start"] : df.iloc[i]["offset_end"]]],
271
+ "offsets": [[df.iloc[i]["offset_start"], df.iloc[i]["offset_end"]]],
272
+ "normalized": [],
273
+ }
274
+ )
275
+ i += 1
276
+ elif df.iloc[i]["span"] is not None:
277
+ ent = df[df["span"] == df.iloc[i]["span"]]
278
+ out["entities"].append(
279
+ {
280
+ "id": f'{uid}-{df.iloc[i]["passage_token_id"]}',
281
+ "type": df.iloc[i]["lab"],
282
+ "text": [text[ent.iloc[0]["offset_start"] : ent.iloc[-1]["offset_end"]]],
283
+ "offsets": [[ent.iloc[0]["offset_start"], ent.iloc[-1]["offset_end"]]],
284
+ "normalized": [],
285
+ }
286
+ )
287
+ i += len(ent)
288
+ else:
289
+ i += 1
290
+
291
+ return out
292
+
293
+
294
+ def _sort_files(filepath):
295
+ doc_ids = os.listdir(filepath)
296
+ doc_ids = [int(doc_ids[i].split(".")[0]) for i in range(len(doc_ids))]
297
+ doc_ids = sorted(doc_ids)
298
+ doc_ids = [f"{doc_ids[i]}.tsv" for i in range(len(doc_ids))]
299
+ return doc_ids
300
+