enoriega commited on
Commit
0a900fd
1 Parent(s): 008dd6b

Added data loading script and sample dataset

Browse files
Files changed (2) hide show
  1. data_files.tar.gz +3 -0
  2. keyword_pubmed.py +135 -0
data_files.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f3fe45f65cb4c3196eba50793c44c55a9d1fb17af8072b01c08b637e54707eb
3
+ size 8785028
keyword_pubmed.py ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """ Loading script for the Keyword PubMed dataset."""
2
+
3
+
4
+ import os
5
+ from pathlib import Path
6
+ import re
7
+
8
+ import datasets
9
+
10
+
11
+ class KeywordPubmedDataset(datasets.GeneratorBasedBuilder):
12
+
13
+ VERSION = datasets.Version("1.0.0")
14
+
15
+
16
+
17
+ BUILDER_CONFIGS = [
18
+ datasets.BuilderConfig(name="sentence", version=VERSION, description="Comprises sentences that contain a keyword"),
19
+ datasets.BuilderConfig(name="document", version=VERSION, description="Contains all the sentences in a document that contains at least a keyword"),
20
+ ]
21
+
22
+ DEFAULT_CONFIG_NAME = "document" # It's not mandatory to have a default configuration. Just use one if it make sense.
23
+
24
+ def _info(self):
25
+ if self.config.name == "sentence":
26
+ features = datasets.Features(
27
+ {
28
+ "sentence": datasets.Value("string"),
29
+ "pmcid": datasets.Value("string"),
30
+ "keyword_rank": datasets.Value("int32"),
31
+ }
32
+ )
33
+ else:
34
+ features = datasets.Features(
35
+ {
36
+ "sentence": datasets.Value("string"),
37
+ "pmcid": datasets.Value("string"),
38
+ "keyword_rank": datasets.Value("int32"),
39
+ }
40
+ )
41
+ return datasets.DatasetInfo(
42
+ # This is the description that will appear on the datasets page.
43
+ description= "Dataset for MLM comprising sentences that contain a keyword relevant to the domain",
44
+ # This defines the different columns of the dataset and their types
45
+ features=features, # Here we define them above because they are different between the two configurations
46
+ # If there's a common (input, target) tuple from the features, uncomment supervised_keys line below and
47
+ # specify them. They'll be used if as_supervised=True in builder.as_dataset.
48
+ # supervised_keys=("sentence", "label"),
49
+ # Homepage of the dataset for documentation
50
+ # homepage=_HOMEPAGE,
51
+ # License for the dataset if available
52
+ # license=_LICENSE,
53
+ # Citation for the dataset
54
+ # citation=_CITATION,
55
+ )
56
+
57
+ def _split_generators(self, dl_manager):
58
+
59
+
60
+ if self.config.data_dir:
61
+ data_dir = self.config.data_dir
62
+ else:
63
+ data_dir = dl_manager.download_and_extract('data_files.tar.gz')
64
+
65
+ # Load the keywords from the file
66
+ with open(os.path.join(data_dir, 'keywords.txt'), 'r') as f:
67
+ keyword_ranks = {line.strip().split(":")[0].lower():rank for rank, line in enumerate(f)}
68
+ keywords = set(keyword_ranks.keys())
69
+
70
+ return [
71
+ datasets.SplitGenerator(
72
+ name=datasets.Split.TRAIN,
73
+ # These kwargs will be passed to _generate_examples
74
+ gen_kwargs={
75
+ "dirpath": os.path.join(data_dir, "train"),
76
+ "keywords": keywords,
77
+ "ranks": keyword_ranks,
78
+ },
79
+ ),
80
+
81
+ datasets.SplitGenerator(
82
+ name=datasets.Split.VALIDATION,
83
+ # These kwargs will be passed to _generate_examples
84
+ gen_kwargs={
85
+ "dirpath": os.path.join(data_dir, "dev"),
86
+ "keywords": keywords,
87
+ "ranks": keyword_ranks,
88
+ },
89
+ ),
90
+ ]
91
+
92
+ # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
93
+ def _generate_examples(self, dirpath, keywords, ranks):
94
+ item_ix = 0
95
+ for filepath in Path(dirpath).iterdir():
96
+ filepath = Path(filepath)
97
+ if filepath.suffix == ".txt":
98
+ pmcid = filepath.name.split(".")[0]
99
+ with filepath.open(encoding="utf-8") as f:
100
+ for sentence in f:
101
+ sentence = sentence.strip()
102
+ if sentence: # Ignore blanks
103
+ sentence = re.sub("\s+", " ", sentence)
104
+ has_keyword, rank = self._has_keyword(sentence, keywords, ranks)
105
+ if self.config.name == "sentence":
106
+ # Yields examples as (key, example) tuples
107
+ if has_keyword:
108
+ yield item_ix, {
109
+ "sentence": sentence,
110
+ "keyword_rank": rank,
111
+ "pmcid": pmcid
112
+ }
113
+ item_ix += 1
114
+
115
+ else: # Else document
116
+ yield item_ix, {
117
+ "sentence": sentence,
118
+ "keyword_rank": rank,
119
+ "pmcid": pmcid
120
+ }
121
+ item_ix += 1
122
+
123
+ def _has_keyword(self, sentence, keywords, ranks):
124
+ # Lowercase and split the sentence
125
+ words = sentence.lower().split()
126
+ # Check every word until it finds a keyword
127
+ for word in words:
128
+ if word in keywords:
129
+ return True, ranks[word]
130
+ return False, -1
131
+
132
+
133
+ if __name__ == "__main__":
134
+ ds = KeywordPubmedDataset()
135
+ ds.download_and_prepare()