maximedb commited on
Commit
2f216d8
1 Parent(s): 5761450

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +163 -1
README.md CHANGED
@@ -1 +1,163 @@
1
- hello
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ annotations_creators:
3
+ - no-annotation
4
+ language_creators:
5
+ - other
6
+ languages:
7
+ - ca
8
+ - en
9
+ - de
10
+ - es
11
+ - fr
12
+ - ru
13
+ - ja
14
+ - it
15
+ - zh
16
+ - pt
17
+ - nl
18
+ - tr
19
+ - pl
20
+ - vi
21
+ - ar
22
+ - id
23
+ - uk
24
+ - ro
25
+ - 'no'
26
+ - th
27
+ - sv
28
+ - el
29
+ - fi
30
+ - he
31
+ - da
32
+ - cs
33
+ - ko
34
+ - fa
35
+ - hi
36
+ - hu
37
+ - sk
38
+ - lt
39
+ - et
40
+ - hr
41
+ - is
42
+ - lv
43
+ - ms
44
+ - bg
45
+ - sr
46
+ - ca
47
+ licenses:
48
+ - cc0-1.0
49
+ multilinguality:
50
+ - multilingual
51
+ pretty_name: MQA - a Multilingual FAQ and CQA Dataset
52
+ size_categories:
53
+ - unknown
54
+ source_datasets:
55
+ - original
56
+ task_categories:
57
+ - question-answering
58
+ task_ids:
59
+ - multiple-choice-qa
60
+ ---
61
+ # MQA
62
+ MQA is a multilingual corpus of questions and answers parsed from the [Common Crawl](https://commoncrawl.org/). Questions are divided between *Frequently Asked Questions* pages and *Community Question Answering* pages.
63
+ ```
64
+ from datasets import load_dataset
65
+ load_dataset("clips/mqa", language="en")
66
+ {
67
+ "name": "the title of the question (if any)",
68
+ "text": "the body of the question (if any)",
69
+ "answers": [{
70
+ "text": "the text of the answer",
71
+ "is_accepted": "true|false"
72
+ }]
73
+ }
74
+ ```
75
+
76
+ ## Languages
77
+ We collected around 234M pairs of questions and answers in 39 different languages. To download a language specific subset you need to specify the language key as configuration. See below for an example.
78
+ ```
79
+ load_dataset("clips/mqa", language="en") # replace "en" by any language listed below
80
+ ```
81
+
82
+ | Language | Key | Pairs | Pages |
83
+ |------------|-----|-----------|-----------|
84
+ | All | all | 6,346,693 | 1,035,649 |
85
+ | English | en | 3,719,484 | 608,796 |
86
+ | German | de | 829,098 | 111,618 |
87
+ | Spanish | es | 482,818 | 75,489 |
88
+ | French | fr | 351,458 | 56,317 |
89
+ | Italian | it | 155,296 | 24,562 |
90
+ | Dutch | nl | 150,819 | 32,574 |
91
+ | Portuguese | pt | 138,778 | 26,169 |
92
+ | Turkish | tr | 102,373 | 19,002 |
93
+ | Russian | ru | 91,771 | 22,643 |
94
+ | Polish | pl | 65,182 | 10,695 |
95
+ | Indonesian | id | 45,839 | 7,910 |
96
+ | Norwegian | no | 37,711 | 5,143 |
97
+ | Swedish | sv | 37,003 | 5,270 |
98
+ | Danish | da | 32,655 | 5,279 |
99
+ | Vietnamese | vi | 27,157 | 5,261 |
100
+ | Finnish | fi | 20,485 | 2,795 |
101
+ | Romanian | ro | 17,066 | 3,554 |
102
+ | Czech | cs | 16,675 | 2,568 |
103
+ | Hebrew | he | 11,212 | 1,921 |
104
+ | Hungarian | hu | 8,598 | 1,264 |
105
+ | Croatian | hr | 5,215 | 819 |
106
+
107
+ ## Data Fields
108
+ #### Nested (per page - default)
109
+ The data is organized by page. Each page contains a list of questions and answers.
110
+ - **id**
111
+ - **language**
112
+ - **num_pairs**: the number of FAQs on the page
113
+ - **domain**: source web domain of the FAQs
114
+ - **qa_pairs**: a list of questions and answers
115
+ - **question**
116
+ - **answer**
117
+ - **language**
118
+
119
+ #### Flattened
120
+ The data is organized by pair (i.e. pages are flattened). You can access the flat version of any language by appending `_flat` to the configuration (e.g. `en_flat`). The data will be returned pair-by-pair instead of page-by-page.
121
+ - **domain_id**
122
+ - **pair_id**
123
+ - **language**
124
+ - **domain**: source web domain of the FAQs
125
+ - **question**
126
+ - **answer**
127
+
128
+ ## Source Data
129
+
130
+ This section was adapted from the source data description of [OSCAR](https://huggingface.co/datasets/oscar#source-data)
131
+
132
+ Common Crawl is a non-profit foundation which produces and maintains an open repository of web crawled data that is both accessible and analysable. Common Crawl's complete web archive consists of petabytes of data collected over 8 years of web crawling. The repository contains raw web page HTML data (WARC files), metdata extracts (WAT files) and plain text extracts (WET files). The organisation's crawlers has always respected nofollow and robots.txt policies.
133
+
134
+ To construct MFAQ, the WARC files of Common Crawl were used. We looked for `FAQPage` markup in the HTML and subsequently parsed the `FAQItem` from the page.
135
+
136
+ ## People
137
+ This model was developed by [Maxime De Bruyn](https://www.linkedin.com/in/maximedebruyn/), Ehsan Lotfi, Jeska Buhmann and Walter Daelemans.
138
+
139
+ ## Licensing Information
140
+ ```
141
+ These data are released under this licensing scheme.
142
+ We do not own any of the text from which these data has been extracted.
143
+ We license the actual packaging of these data under the Creative Commons CC0 license ("no rights reserved") http://creativecommons.org/publicdomain/zero/1.0/
144
+
145
+ Should you consider that our data contains material that is owned by you and should therefore not be reproduced here, please:
146
+ * Clearly identify yourself, with detailed contact data such as an address, telephone number or email address at which you can be contacted.
147
+ * Clearly identify the copyrighted work claimed to be infringed.
148
+ * Clearly identify the material that is claimed to be infringing and information reasonably sufficient to allow us to locate the material.
149
+
150
+ We will comply to legitimate requests by removing the affected sources from the next release of the corpus.
151
+ ```
152
+
153
+ ## Citation information
154
+ ```
155
+ @misc{debruyn2021mfaq,
156
+ title={MFAQ: a Multilingual FAQ Dataset},
157
+ author={Maxime {De Bruyn} and Ehsan Lotfi and Jeska Buhmann and Walter Daelemans},
158
+ year={2021},
159
+ eprint={2109.12870},
160
+ archivePrefix={arXiv},
161
+ primaryClass={cs.CL}
162
+ }
163
+ ```