Commit
•
92f7ed7
1
Parent(s):
86db073
Delete loading script
Browse files- ms_marco.py +0 -204
ms_marco.py
DELETED
@@ -1,204 +0,0 @@
|
|
1 |
-
# coding=utf-8
|
2 |
-
# Copyright 2020 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors.
|
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 |
-
"""MS MARCO dataset."""
|
18 |
-
|
19 |
-
|
20 |
-
import json
|
21 |
-
|
22 |
-
import datasets
|
23 |
-
|
24 |
-
|
25 |
-
_CITATION = """
|
26 |
-
@article{DBLP:journals/corr/NguyenRSGTMD16,
|
27 |
-
author = {Tri Nguyen and
|
28 |
-
Mir Rosenberg and
|
29 |
-
Xia Song and
|
30 |
-
Jianfeng Gao and
|
31 |
-
Saurabh Tiwary and
|
32 |
-
Rangan Majumder and
|
33 |
-
Li Deng},
|
34 |
-
title = {{MS} {MARCO:} {A} Human Generated MAchine Reading COmprehension Dataset},
|
35 |
-
journal = {CoRR},
|
36 |
-
volume = {abs/1611.09268},
|
37 |
-
year = {2016},
|
38 |
-
url = {http://arxiv.org/abs/1611.09268},
|
39 |
-
archivePrefix = {arXiv},
|
40 |
-
eprint = {1611.09268},
|
41 |
-
timestamp = {Mon, 13 Aug 2018 16:49:03 +0200},
|
42 |
-
biburl = {https://dblp.org/rec/journals/corr/NguyenRSGTMD16.bib},
|
43 |
-
bibsource = {dblp computer science bibliography, https://dblp.org}
|
44 |
-
}
|
45 |
-
}
|
46 |
-
"""
|
47 |
-
|
48 |
-
_DESCRIPTION = """
|
49 |
-
Starting with a paper released at NIPS 2016, MS MARCO is a collection of datasets focused on deep learning in search.
|
50 |
-
|
51 |
-
The first dataset was a question answering dataset featuring 100,000 real Bing questions and a human generated answer.
|
52 |
-
Since then we released a 1,000,000 question dataset, a natural langauge generation dataset, a passage ranking dataset,
|
53 |
-
keyphrase extraction dataset, crawling dataset, and a conversational search.
|
54 |
-
|
55 |
-
There have been 277 submissions. 20 KeyPhrase Extraction submissions, 87 passage ranking submissions, 0 document ranking
|
56 |
-
submissions, 73 QnA V2 submissions, 82 NLGEN submisions, and 15 QnA V1 submissions
|
57 |
-
|
58 |
-
This data comes in three tasks/forms: Original QnA dataset(v1.1), Question Answering(v2.1), Natural Language Generation(v2.1).
|
59 |
-
|
60 |
-
The original question answering datset featured 100,000 examples and was released in 2016. Leaderboard is now closed but data is availible below.
|
61 |
-
|
62 |
-
The current competitive tasks are Question Answering and Natural Language Generation. Question Answering features over 1,000,000 queries and
|
63 |
-
is much like the original QnA dataset but bigger and with higher quality. The Natural Language Generation dataset features 180,000 examples and
|
64 |
-
builds upon the QnA dataset to deliver answers that could be spoken by a smart speaker.
|
65 |
-
|
66 |
-
"""
|
67 |
-
_V2_URLS = {
|
68 |
-
"train": "https://msmarco.blob.core.windows.net/msmarco/train_v2.1.json.gz",
|
69 |
-
"dev": "https://msmarco.blob.core.windows.net/msmarco/dev_v2.1.json.gz",
|
70 |
-
"test": "https://msmarco.blob.core.windows.net/msmarco/eval_v2.1_public.json.gz",
|
71 |
-
}
|
72 |
-
|
73 |
-
_V1_URLS = {
|
74 |
-
"train": "https://msmarco.blob.core.windows.net/msmsarcov1/train_v1.1.json.gz",
|
75 |
-
"dev": "https://msmarco.blob.core.windows.net/msmsarcov1/dev_v1.1.json.gz",
|
76 |
-
"test": "https://msmarco.blob.core.windows.net/msmsarcov1/test_hidden_v1.1.json",
|
77 |
-
}
|
78 |
-
|
79 |
-
|
80 |
-
class MsMarcoConfig(datasets.BuilderConfig):
|
81 |
-
"""BuilderConfig for MS MARCO."""
|
82 |
-
|
83 |
-
def __init__(self, **kwargs):
|
84 |
-
"""BuilderConfig for MS MARCO
|
85 |
-
|
86 |
-
Args:
|
87 |
-
**kwargs: keyword arguments forwarded to super.
|
88 |
-
"""
|
89 |
-
super(MsMarcoConfig, self).__init__(**kwargs)
|
90 |
-
|
91 |
-
|
92 |
-
class MsMarco(datasets.GeneratorBasedBuilder):
|
93 |
-
|
94 |
-
BUILDER_CONFIGS = [
|
95 |
-
MsMarcoConfig(
|
96 |
-
name="v1.1",
|
97 |
-
description="""version v1.1""",
|
98 |
-
version=datasets.Version("1.1.0", ""),
|
99 |
-
),
|
100 |
-
MsMarcoConfig(
|
101 |
-
name="v2.1",
|
102 |
-
description="""version v2.1""",
|
103 |
-
version=datasets.Version("2.1.0", ""),
|
104 |
-
),
|
105 |
-
]
|
106 |
-
|
107 |
-
def _info(self):
|
108 |
-
return datasets.DatasetInfo(
|
109 |
-
description=_DESCRIPTION + "\n" + self.config.description,
|
110 |
-
features=datasets.Features(
|
111 |
-
{
|
112 |
-
"answers": datasets.features.Sequence(datasets.Value("string")),
|
113 |
-
"passages": datasets.features.Sequence(
|
114 |
-
{
|
115 |
-
"is_selected": datasets.Value("int32"),
|
116 |
-
"passage_text": datasets.Value("string"),
|
117 |
-
"url": datasets.Value("string"),
|
118 |
-
}
|
119 |
-
),
|
120 |
-
"query": datasets.Value("string"),
|
121 |
-
"query_id": datasets.Value("int32"),
|
122 |
-
"query_type": datasets.Value("string"),
|
123 |
-
"wellFormedAnswers": datasets.features.Sequence(datasets.Value("string")),
|
124 |
-
}
|
125 |
-
),
|
126 |
-
homepage="https://microsoft.github.io/msmarco/",
|
127 |
-
citation=_CITATION,
|
128 |
-
)
|
129 |
-
|
130 |
-
def _split_generators(self, dl_manager):
|
131 |
-
"""Returns SplitGenerators."""
|
132 |
-
if self.config.name == "v2.1":
|
133 |
-
dl_path = dl_manager.download_and_extract(_V2_URLS)
|
134 |
-
else:
|
135 |
-
dl_path = dl_manager.download_and_extract(_V1_URLS)
|
136 |
-
return [
|
137 |
-
datasets.SplitGenerator(
|
138 |
-
name=datasets.Split.VALIDATION,
|
139 |
-
gen_kwargs={"filepath": dl_path["dev"]},
|
140 |
-
),
|
141 |
-
datasets.SplitGenerator(
|
142 |
-
name=datasets.Split.TRAIN,
|
143 |
-
gen_kwargs={"filepath": dl_path["train"]},
|
144 |
-
),
|
145 |
-
datasets.SplitGenerator(
|
146 |
-
name=datasets.Split.TEST,
|
147 |
-
gen_kwargs={"filepath": dl_path["test"]},
|
148 |
-
),
|
149 |
-
]
|
150 |
-
|
151 |
-
def _generate_examples(self, filepath):
|
152 |
-
"""Yields examples."""
|
153 |
-
with open(filepath, encoding="utf-8") as f:
|
154 |
-
if self.config.name == "v2.1":
|
155 |
-
data = json.load(f)
|
156 |
-
questions = data["query"]
|
157 |
-
answers = data.get("answers", {})
|
158 |
-
passages = data["passages"]
|
159 |
-
query_ids = data["query_id"]
|
160 |
-
query_types = data["query_type"]
|
161 |
-
wellFormedAnswers = data.get("wellFormedAnswers", {})
|
162 |
-
for key in questions:
|
163 |
-
|
164 |
-
is_selected = [passage.get("is_selected", -1) for passage in passages[key]]
|
165 |
-
passage_text = [passage["passage_text"] for passage in passages[key]]
|
166 |
-
urls = [passage["url"] for passage in passages[key]]
|
167 |
-
question = questions[key]
|
168 |
-
answer = answers.get(key, [])
|
169 |
-
query_id = query_ids[key]
|
170 |
-
query_type = query_types[key]
|
171 |
-
wellFormedAnswer = wellFormedAnswers.get(key, [])
|
172 |
-
if wellFormedAnswer == "[]":
|
173 |
-
wellFormedAnswer = []
|
174 |
-
yield query_id, {
|
175 |
-
"answers": answer,
|
176 |
-
"passages": {"is_selected": is_selected, "passage_text": passage_text, "url": urls},
|
177 |
-
"query": question,
|
178 |
-
"query_id": query_id,
|
179 |
-
"query_type": query_type,
|
180 |
-
"wellFormedAnswers": wellFormedAnswer,
|
181 |
-
}
|
182 |
-
if self.config.name == "v1.1":
|
183 |
-
for row in f:
|
184 |
-
data = json.loads(row)
|
185 |
-
question = data["query"]
|
186 |
-
answer = data.get("answers", [])
|
187 |
-
passages = data["passages"]
|
188 |
-
query_id = data["query_id"]
|
189 |
-
query_type = data["query_type"]
|
190 |
-
wellFormedAnswer = data.get("wellFormedAnswers", [])
|
191 |
-
|
192 |
-
is_selected = [passage.get("is_selected", -1) for passage in passages]
|
193 |
-
passage_text = [passage["passage_text"] for passage in passages]
|
194 |
-
urls = [passage["url"] for passage in passages]
|
195 |
-
if wellFormedAnswer == "[]":
|
196 |
-
wellFormedAnswer = []
|
197 |
-
yield query_id, {
|
198 |
-
"answers": answer,
|
199 |
-
"passages": {"is_selected": is_selected, "passage_text": passage_text, "url": urls},
|
200 |
-
"query": question,
|
201 |
-
"query_id": query_id,
|
202 |
-
"query_type": query_type,
|
203 |
-
"wellFormedAnswers": wellFormedAnswer,
|
204 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|