crystina-z
commited on
Commit
•
17f4a45
1
Parent(s):
15c5596
Update miracl.py
Browse files
miracl.py
CHANGED
@@ -0,0 +1,142 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
# Lint as: python3
|
17 |
+
|
18 |
+
import json
|
19 |
+
import datasets
|
20 |
+
from collections import defaultdict
|
21 |
+
from dataclasses import dataclass
|
22 |
+
|
23 |
+
_CITATION = '''
|
24 |
+
'''
|
25 |
+
|
26 |
+
languages = ['ar', 'bn', 'en', 'es', 'fa', 'fi', 'fr', 'hi', 'id', 'ja', 'ko', 'ru', 'sw', 'te', 'th', 'zh']
|
27 |
+
non_surprise_languages = languages
|
28 |
+
|
29 |
+
_DESCRIPTION = 'dataset load script for MIRACL'
|
30 |
+
|
31 |
+
_DATASET_URLS = {
|
32 |
+
lang: {
|
33 |
+
'train': [
|
34 |
+
f'https://huggingface.co/datasets/miracl/miracl/resolve/main/miracl-v1.0-{lang}/topics/{lang}.topic.train.tsv',
|
35 |
+
f'https://huggingface.co/datasets/miracl/miracl/resolve/main/miracl-v1.0-{lang}/qrels/{lang}.qrels.train.tsv',
|
36 |
+
],
|
37 |
+
'dev': [
|
38 |
+
f'https://huggingface.co/datasets/miracl/miracl/resolve/main/miracl-v1.0-{lang}/topics/{lang}.topic.dev.tsv',
|
39 |
+
f'https://huggingface.co/datasets/miracl/miracl/resolve/main/miracl-v1.0-{lang}/qrels/{lang}.qrels.dev.tsv',
|
40 |
+
]
|
41 |
+
} for lang in languages
|
42 |
+
}
|
43 |
+
|
44 |
+
|
45 |
+
def load_topic(fn):
|
46 |
+
qid2topic = {}
|
47 |
+
with open(fn, encoding="utf-8") as f:
|
48 |
+
for line in f:
|
49 |
+
qid, topic = line.strip().split('\t')
|
50 |
+
qid2topic[qid] = topic
|
51 |
+
return qid2topic
|
52 |
+
|
53 |
+
|
54 |
+
def load_qrels(fn):
|
55 |
+
qrels = defaultdict(dict)
|
56 |
+
with open(fn, encoding="utf-8") as f:
|
57 |
+
for line in f:
|
58 |
+
qid, _, docid, rel = line.strip().split('\t')
|
59 |
+
qrels[qid][docid] = int(rel)
|
60 |
+
return qrels
|
61 |
+
|
62 |
+
|
63 |
+
class MIRACL(datasets.GeneratorBasedBuilder):
|
64 |
+
BUILDER_CONFIGS = [datasets.BuilderConfig(
|
65 |
+
version=datasets.Version('1.0.0'),
|
66 |
+
name=lang, description=f'MIRACL dataset in language {lang}.'
|
67 |
+
) for lang in languages
|
68 |
+
]
|
69 |
+
|
70 |
+
def _info(self):
|
71 |
+
features = datasets.Features({
|
72 |
+
'query_id': datasets.Value('string'),
|
73 |
+
'query': datasets.Value('string'),
|
74 |
+
|
75 |
+
'positive_passages': [{
|
76 |
+
'docid': datasets.Value('string'),
|
77 |
+
'text': datasets.Value('string'), 'title': datasets.Value('string')
|
78 |
+
}],
|
79 |
+
'negative_passages': [{
|
80 |
+
'docid': datasets.Value('string'),
|
81 |
+
'text': datasets.Value('string'), 'title': datasets.Value('string'),
|
82 |
+
}],
|
83 |
+
})
|
84 |
+
|
85 |
+
return datasets.DatasetInfo(
|
86 |
+
# This is the description that will appear on the datasets page.
|
87 |
+
description=_DESCRIPTION,
|
88 |
+
# This defines the different columns of the dataset and their types
|
89 |
+
features=features, # Here we define them above because they are different between the two configurations
|
90 |
+
supervised_keys=None,
|
91 |
+
# Homepage of the dataset for documentation
|
92 |
+
homepage='https://project-miracl.github.io',
|
93 |
+
# License for the dataset if available
|
94 |
+
license='',
|
95 |
+
# Citation for the dataset
|
96 |
+
citation=_CITATION,
|
97 |
+
)
|
98 |
+
|
99 |
+
def _split_generators(self, dl_manager):
|
100 |
+
lang = self.config.name
|
101 |
+
downloaded_files = dl_manager.download_and_extract(_DATASET_URLS[lang])
|
102 |
+
|
103 |
+
splits = [
|
104 |
+
datasets.SplitGenerator(
|
105 |
+
name='train',
|
106 |
+
gen_kwargs={
|
107 |
+
'filepaths': downloaded_files['train'],
|
108 |
+
},
|
109 |
+
),
|
110 |
+
datasets.SplitGenerator(
|
111 |
+
name='dev',
|
112 |
+
gen_kwargs={
|
113 |
+
'filepaths': downloaded_files['dev'],
|
114 |
+
},
|
115 |
+
),
|
116 |
+
]
|
117 |
+
return splits
|
118 |
+
|
119 |
+
def _generate_examples(self, filepaths):
|
120 |
+
lang = self.config.name
|
121 |
+
miracl_corpus = datasets.load_dataset('miracl/miracl-corpus', lang)['train']
|
122 |
+
docid2doc = {doc['docid']: (doc['title'], doc['text']) for doc in miracl_corpus}
|
123 |
+
|
124 |
+
topic_fn, qrel_fn = filepaths
|
125 |
+
qid2topic = load_topic(topic_fn)
|
126 |
+
qrels = load_qrels(qrel_fn)
|
127 |
+
for qid in qid2topic:
|
128 |
+
data = {}
|
129 |
+
data['query_id'] = qid
|
130 |
+
data['query'] = qid2topic[qid]
|
131 |
+
|
132 |
+
pos_docids = [docid for docid, rel in qrels[qid].items() if rel == 1]
|
133 |
+
neg_docids = [docid for docid, rel in qrels[qid].items() if rel == 0]
|
134 |
+
data['positive_passages'] = [{
|
135 |
+
'docid': docid,
|
136 |
+
**dict(zip(['title', 'text'], docid2doc[docid]))
|
137 |
+
} for docid in pos_docids]
|
138 |
+
data['negative_passages'] = [{
|
139 |
+
'docid': docid,
|
140 |
+
**dict(zip(['title', 'text'], docid2doc[docid]))
|
141 |
+
} for docid in neg_docids]
|
142 |
+
yield qid, data
|