Datasets:

Languages:
English
ArXiv:
License:
gabrielaltay commited on
Commit
6cc92f0
1 Parent(s): 786333b

upload hubscripts/pmc_patients_hub.py to hub from bigbio repo

Browse files
Files changed (1) hide show
  1. pmc_patients.py +207 -0
pmc_patients.py ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ """
17
+ PPS dataset is a list of triplets. Each entry is in format (patient_uid_1, patient_uid_2, similarity)
18
+ where similarity has three values:0, 1, 2, indicating corresponding similarity.
19
+ """
20
+
21
+ import json
22
+ import os
23
+ from typing import Dict, List, Tuple
24
+
25
+ import datasets
26
+ import pandas as pd
27
+
28
+ from .bigbiohub import pairs_features
29
+ from .bigbiohub import BigBioConfig
30
+ from .bigbiohub import Tasks
31
+
32
+ _LANGUAGES = ['English']
33
+ _PUBMED = True
34
+ _LOCAL = False
35
+ _CITATION = """\
36
+ @misc{zhao2022pmcpatients,
37
+ title={PMC-Patients: A Large-scale Dataset of Patient Notes and Relations Extracted from Case
38
+ Reports in PubMed Central},
39
+ author={Zhengyun Zhao and Qiao Jin and Sheng Yu},
40
+ year={2022},
41
+ eprint={2202.13876},
42
+ archivePrefix={arXiv},
43
+ primaryClass={cs.CL}
44
+ }"""
45
+
46
+ _DATASETNAME = "pmc_patients"
47
+ _DISPLAYNAME = "PMC-Patients"
48
+
49
+ _DESCRIPTION = """\
50
+ This dataset is used for calculating the similarity between two patient descriptions.
51
+ """
52
+
53
+ _HOMEPAGE = "https://github.com/zhao-zy15/PMC-Patients"
54
+
55
+ _LICENSE = 'Creative Commons Attribution Non Commercial Share Alike 4.0 International'
56
+
57
+ _URLS = {
58
+ _DATASETNAME: "https://drive.google.com/u/0/uc?id=1vFCLy_CF8fxPDZvDtHPR6Dl6x9l0TyvW&export=download",
59
+ }
60
+
61
+ _SUPPORTED_TASKS = [Tasks.SEMANTIC_SIMILARITY]
62
+
63
+ _SOURCE_VERSION = "1.2.0"
64
+
65
+ _BIGBIO_VERSION = "1.0.0"
66
+
67
+
68
+ class PMCPatientsDataset(datasets.GeneratorBasedBuilder):
69
+ """PPS dataset is a list of triplets.
70
+ Each entry is in format (patient_uid_1, patient_uid_2, similarity) and their
71
+ respective texts.
72
+ where similarity has three values:0, 1, 2, indicating corresponding similarity.
73
+ """
74
+
75
+ SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
76
+ BIGBIO_VERSION = datasets.Version(_BIGBIO_VERSION)
77
+
78
+ BUILDER_CONFIGS = [
79
+ BigBioConfig(
80
+ name="pmc_patients_source",
81
+ version=SOURCE_VERSION,
82
+ description="pmc_patients source schema",
83
+ schema="source",
84
+ subset_id="pmc_patients",
85
+ ),
86
+ BigBioConfig(
87
+ name="pmc_patients_bigbio_pairs",
88
+ version=BIGBIO_VERSION,
89
+ description="pmc_patients BigBio schema",
90
+ schema="bigbio_pairs",
91
+ subset_id="pmc_patients",
92
+ ),
93
+ ]
94
+
95
+ DEFAULT_CONFIG_NAME = "pmc_patients_source"
96
+
97
+ def _info(self) -> datasets.DatasetInfo:
98
+ if self.config.schema == "source":
99
+ features = datasets.Features(
100
+ {
101
+ "id": datasets.Value("string"),
102
+ "id_text1": datasets.Value("string"),
103
+ "id_text2": datasets.Value("string"),
104
+ "label": datasets.Value("int8"),
105
+ }
106
+ )
107
+
108
+ elif self.config.schema == "bigbio_pairs":
109
+ features = pairs_features
110
+
111
+ return datasets.DatasetInfo(
112
+ description=_DESCRIPTION,
113
+ features=features,
114
+ homepage=_HOMEPAGE,
115
+ license=str(_LICENSE),
116
+ citation=_CITATION,
117
+ )
118
+
119
+ def _split_generators(self, dl_manager) -> List[datasets.SplitGenerator]:
120
+ """Returns SplitGenerators."""
121
+ urls = _URLS[_DATASETNAME]
122
+ data_dir = dl_manager.download_and_extract(urls)
123
+ return [
124
+ datasets.SplitGenerator(
125
+ name=datasets.Split.TRAIN,
126
+ gen_kwargs={
127
+ "filepath": os.path.join(
128
+ data_dir,
129
+ "datasets/task_2_patient2patient_similarity/PPS_train.json",
130
+ ),
131
+ "split": "train",
132
+ "data_dir": data_dir,
133
+ },
134
+ ),
135
+ datasets.SplitGenerator(
136
+ name=datasets.Split.TEST,
137
+ gen_kwargs={
138
+ "filepath": os.path.join(
139
+ data_dir,
140
+ "datasets/task_2_patient2patient_similarity/PPS_test.json",
141
+ ),
142
+ "split": "test",
143
+ "data_dir": data_dir,
144
+ },
145
+ ),
146
+ datasets.SplitGenerator(
147
+ name=datasets.Split.VALIDATION,
148
+ gen_kwargs={
149
+ "filepath": os.path.join(
150
+ data_dir,
151
+ "datasets/task_2_patient2patient_similarity/PPS_dev.json",
152
+ ),
153
+ "split": "dev",
154
+ "data_dir": data_dir,
155
+ },
156
+ ),
157
+ ]
158
+
159
+ def _generate_examples(
160
+ self, filepath, split: str, data_dir: str
161
+ ) -> Tuple[int, Dict]:
162
+ """Yields examples as (key, example) tuples."""
163
+
164
+ uid = 0
165
+
166
+ def lookup_text(patient_uid: str, df: pd.DataFrame) -> str:
167
+ try:
168
+ return df.loc[patient_uid]["patient"]
169
+ except KeyError:
170
+ return ""
171
+
172
+ with open(filepath, "r") as j:
173
+ ret_file = json.load(j)
174
+
175
+ if self.config.schema == "source":
176
+
177
+ for key, (id1, id2, label) in enumerate(ret_file):
178
+ feature_dict = {
179
+ "id": uid,
180
+ "id_text1": id1,
181
+ "id_text2": id2,
182
+ "label": label,
183
+ }
184
+ uid += 1
185
+ yield key, feature_dict
186
+
187
+ elif self.config.schema == "bigbio_pairs":
188
+ source_files = os.path.join(data_dir, f"datasets/PMC-Patients_{split}.json")
189
+ src_frame = pd.read_json(source_files, encoding="utf8").set_index(
190
+ "patient_uid"
191
+ )
192
+ for key, (id1, id2, label) in enumerate(ret_file):
193
+ text_1 = lookup_text(id1, src_frame)
194
+ text_2 = lookup_text(id2, src_frame)
195
+ # test/dev splits are faulty and may not contain the patient_uid
196
+ # if any of the lookup texts are empty skip the sample
197
+ if text_1 == "" or text_2 == "":
198
+ continue
199
+ feature_dict = {
200
+ "id": uid,
201
+ "document_id": "NULL",
202
+ "text_1": text_1,
203
+ "text_2": text_2,
204
+ "label": label,
205
+ }
206
+ uid += 1
207
+ yield key, feature_dict