quentinbrabant
commited on
Commit
•
7079747
1
Parent(s):
c03b746
Upload CoQAR.py
Browse files
CoQAR.py
ADDED
@@ -0,0 +1,174 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
|
2 |
+
#
|
3 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
4 |
+
# you may not use this file except in compliance with the License.
|
5 |
+
# You may obtain a copy of the License at
|
6 |
+
#
|
7 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
8 |
+
#
|
9 |
+
# Unless required by applicable law or agreed to in writing, software
|
10 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
11 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12 |
+
# See the License for the specific language governing permissions and
|
13 |
+
# limitations under the License.
|
14 |
+
|
15 |
+
"""
|
16 |
+
CoQAR is a corpus containing 4.5K conversations from the open-source dataset [Conversational Question-Answering dataset CoQA](https://stanfordnlp.github.io/coqa/), for a total of 53K follow-up question-answer pairs.
|
17 |
+
In CoQAR each original question was manually annotated with at least 2 at most 3 out-of-context rewritings.
|
18 |
+
COQAR can be used for (at least) three NLP tasks: question paraphrasing, question rewriting and conversational question answering.
|
19 |
+
|
20 |
+
We annotated each original question of CoQA with at least 2 at most 3 out-of-context rewritings.
|
21 |
+
|
22 |
+
![image](https://user-images.githubusercontent.com/52821991/165952155-822ce743-791d-46c8-8705-0937a69df933.png)
|
23 |
+
|
24 |
+
The annotations are published under the licence CC-BY-SA 4.0.
|
25 |
+
The original content of the dataset CoQA is under the distinct licences described below.
|
26 |
+
|
27 |
+
The corpus CoQA contains passages from seven domains, which are public under the following licenses:
|
28 |
+
- Literature and Wikipedia passages are shared under CC BY-SA 4.0 license.
|
29 |
+
- Children's stories are collected from MCTest which comes with MSR-LA license.
|
30 |
+
- Middle/High school exam passages are collected from RACE which comes with its own license.
|
31 |
+
- News passages are collected from the DeepMind CNN dataset which comes with Apache license (see [K. M. Hermann, T. Kočiský and E. Grefenstette, L. Espeholt, W. Kay, M. Suleyman, P. Blunsom, Teaching Machines to Read and Comprehend. Advances in Neural Information Processing Systems (NIPS), 2015](http://arxiv.org/abs/1506.03340)).
|
32 |
+
"""
|
33 |
+
|
34 |
+
|
35 |
+
import csv
|
36 |
+
import json
|
37 |
+
import os
|
38 |
+
|
39 |
+
import datasets
|
40 |
+
|
41 |
+
_CITATION = """\
|
42 |
+
@inproceedings{brabant-etal-2022-coqar,
|
43 |
+
title = "{C}o{QAR}: Question Rewriting on {C}o{QA}",
|
44 |
+
author = "Brabant, Quentin and
|
45 |
+
Lecorv{\'e}, Gw{\'e}nol{\'e} and
|
46 |
+
Rojas Barahona, Lina M.",
|
47 |
+
booktitle = "Proceedings of the Thirteenth Language Resources and Evaluation Conference",
|
48 |
+
month = jun,
|
49 |
+
year = "2022",
|
50 |
+
address = "Marseille, France",
|
51 |
+
publisher = "European Language Resources Association",
|
52 |
+
url = "https://aclanthology.org/2022.lrec-1.13",
|
53 |
+
pages = "119--126"
|
54 |
+
}
|
55 |
+
"""
|
56 |
+
|
57 |
+
_DESCRIPTION = """\
|
58 |
+
CoQAR is a corpus containing 4.5K conversations from the open-source dataset [Conversational Question-Answering dataset CoQA](https://stanfordnlp.github.io/coqa/), for a total of 53K follow-up question-answer pairs.
|
59 |
+
In CoQAR each original question was manually annotated with at least 2 at most 3 out-of-context rewritings.
|
60 |
+
COQAR can be used for (at least) three NLP tasks: question paraphrasing, question rewriting and conversational question answering.
|
61 |
+
|
62 |
+
We annotated each original question of CoQA with at least 2 at most 3 out-of-context rewritings.
|
63 |
+
|
64 |
+
![image](https://user-images.githubusercontent.com/52821991/165952155-822ce743-791d-46c8-8705-0937a69df933.png)
|
65 |
+
|
66 |
+
The annotations are published under the licence CC-BY-SA 4.0.
|
67 |
+
The original content of the dataset CoQA is under the distinct licences described below.
|
68 |
+
|
69 |
+
The corpus CoQA contains passages from seven domains, which are public under the following licenses:
|
70 |
+
- Literature and Wikipedia passages are shared under CC BY-SA 4.0 license.
|
71 |
+
- Children's stories are collected from MCTest which comes with MSR-LA license.
|
72 |
+
- Middle/High school exam passages are collected from RACE which comes with its own license.
|
73 |
+
- News passages are collected from the DeepMind CNN dataset which comes with Apache license (see [K. M. Hermann, T. Kočiský and E. Grefenstette, L. Espeholt, W. Kay, M. Suleyman, P. Blunsom, Teaching Machines to Read and Comprehend. Advances in Neural Information Processing Systems (NIPS), 2015](http://arxiv.org/abs/1506.03340)).
|
74 |
+
"""
|
75 |
+
|
76 |
+
_HOMEPAGE = "https://github.com/Orange-OpenSource/COQAR/"
|
77 |
+
|
78 |
+
_LICENSE = """
|
79 |
+
- Annotations, litterature and Wikipedia passages: licence CC-BY-SA 4.0.
|
80 |
+
- Children's stories are from MCTest (MSR-LA license).
|
81 |
+
- Exam passages come from RACE which has its own license.
|
82 |
+
- News passages are from the DeepMind CNN dataset (Apache license).
|
83 |
+
"""
|
84 |
+
|
85 |
+
_URLS = {
|
86 |
+
"train": "https://raw.githubusercontent.com/Orange-OpenSource/COQAR/master/data/CoQAR/train/coqar-train-v1.0.json",
|
87 |
+
"dev": "https://raw.githubusercontent.com/Orange-OpenSource/COQAR/master/data/CoQAR/dev/coqar-dev-v1.0.json"
|
88 |
+
}
|
89 |
+
|
90 |
+
|
91 |
+
class CoQAR(datasets.GeneratorBasedBuilder):
|
92 |
+
"""
|
93 |
+
CoQAR is a corpus containing 4.5K conversations from the open-source dataset [Conversational Question-Answering dataset CoQA](https://stanfordnlp.github.io/coqa/), for a total of 53K follow-up question-answer pairs.
|
94 |
+
In CoQAR each original question was manually annotated with at least 2 at most 3 out-of-context rewritings.
|
95 |
+
COQAR can be used for (at least) three NLP tasks: question paraphrasing, question rewriting and conversational question answering.
|
96 |
+
"""
|
97 |
+
|
98 |
+
VERSION = datasets.Version("1.1.0")
|
99 |
+
|
100 |
+
def _info(self):
|
101 |
+
features = datasets.Features(
|
102 |
+
{
|
103 |
+
'conversation_id' : datasets.Value("string"),
|
104 |
+
'turn_id': datasets.Value("int16"),
|
105 |
+
'original_question' : datasets.Value("string"),
|
106 |
+
'question_paraphrases' : datasets.Sequence(feature=datasets.Value("string")),
|
107 |
+
'answer' : datasets.Value("string"),
|
108 |
+
'answer_span_start' : datasets.Value("int32"),
|
109 |
+
'answer_span_end' : datasets.Value("int32"),
|
110 |
+
'answer_span_text' : datasets.Value("string"),
|
111 |
+
'conversation_history' : datasets.Sequence(feature=datasets.Value("string")),
|
112 |
+
'file_name' : datasets.Value("string"),
|
113 |
+
'story': datasets.Value("string"),
|
114 |
+
'name': datasets.Value("string"),
|
115 |
+
}
|
116 |
+
)
|
117 |
+
|
118 |
+
return datasets.DatasetInfo(
|
119 |
+
# This is the description that will appear on the datasets page.
|
120 |
+
description=_DESCRIPTION,
|
121 |
+
# This defines the different columns of the dataset and their types
|
122 |
+
features=features,
|
123 |
+
homepage=_HOMEPAGE,
|
124 |
+
# License for the dataset if available
|
125 |
+
license=_LICENSE,
|
126 |
+
# Citation for the dataset
|
127 |
+
citation=_CITATION,
|
128 |
+
)
|
129 |
+
|
130 |
+
def _split_generators(self, dl_manager):
|
131 |
+
data_dir = dl_manager.download_and_extract(_URLS)
|
132 |
+
|
133 |
+
return [
|
134 |
+
datasets.SplitGenerator(
|
135 |
+
name=datasets.Split.TRAIN,
|
136 |
+
gen_kwargs={
|
137 |
+
"filepath": data_dir['train'],
|
138 |
+
"split": "train",
|
139 |
+
},
|
140 |
+
),
|
141 |
+
datasets.SplitGenerator(
|
142 |
+
name=datasets.Split.VALIDATION,
|
143 |
+
gen_kwargs={
|
144 |
+
"filepath": data_dir['dev'],
|
145 |
+
"split": "dev",
|
146 |
+
},
|
147 |
+
)
|
148 |
+
]
|
149 |
+
|
150 |
+
# method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
|
151 |
+
def _generate_examples(self, filepath, split):
|
152 |
+
with open(filepath, 'r') as f:
|
153 |
+
dic = json.load(f)
|
154 |
+
i = 0
|
155 |
+
for datum in dic['data']:
|
156 |
+
history = []
|
157 |
+
for question, answer in zip(datum['questions'], datum['answers']):
|
158 |
+
yield i, {
|
159 |
+
'conversation_id' : datum['id'],
|
160 |
+
'turn_id': question['turn_id'],
|
161 |
+
'original_question' :question['input_text'],
|
162 |
+
'question_paraphrases' : question['paraphrase'],
|
163 |
+
'answer' : answer['input_text'],
|
164 |
+
'answer_span_start' : answer['span_start'],
|
165 |
+
'answer_span_end' : answer['span_end'],
|
166 |
+
'answer_span_text' : answer['span_text'],
|
167 |
+
'conversation_history' : list(history),
|
168 |
+
'file_name' : datum['filename'],
|
169 |
+
'story': datum['story'],
|
170 |
+
'name': datum['name']
|
171 |
+
}
|
172 |
+
history.append(question['input_text'])
|
173 |
+
history.append(answer['input_text'])
|
174 |
+
i+=1
|