Datasets:

Modalities:
Text
Formats:
json
ArXiv:
Libraries:
Datasets
pandas
quentinbrabant commited on
Commit
a8dd060
1 Parent(s): f778f52

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +270 -1
README.md CHANGED
@@ -1,3 +1,272 @@
1
  ---
2
- license: cc-by-4.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ configs:
3
+ - config_name: labels
4
+ data_files: data/labels.json
5
+ - config_name: templates
6
+ data_files: data/templates.json
7
+ - config_name: conversations.country
8
+ data_files:
9
+ - path: data/country/test.json
10
+ split: test
11
+ - path: data/country/dev.json
12
+ split: dev
13
+ - path: data/country/train.json
14
+ split: train
15
+ - config_name: conversations.historical_event
16
+ data_files:
17
+ - path: data/historical_event/test.json
18
+ split: test
19
+ - path: data/historical_event/dev.json
20
+ split: dev
21
+ - path: data/historical_event/train.json
22
+ split: train
23
+ - config_name: conversations.food
24
+ data_files:
25
+ - path: data/food/test.json
26
+ split: test
27
+ - path: data/food/dev.json
28
+ split: dev
29
+ - path: data/food/train.json
30
+ split: train
31
+ - config_name: conversations.space_object
32
+ data_files:
33
+ - path: data/space_object/test.json
34
+ split: test
35
+ - config_name: conversations.with_unseen_properties
36
+ data_files:
37
+ - path: data/with_unseen_properties/test.json
38
+ split: test
39
+ - config_name: conversations.taxon
40
+ data_files:
41
+ - path: data/taxon/test.json
42
+ split: test
43
+ - config_name: conversations.person
44
+ data_files:
45
+ - path: data/person/test.json
46
+ split: test
47
+ - path: data/person/dev.json
48
+ split: dev
49
+ - path: data/person/train.json
50
+ split: train
51
+ - config_name: conversations.ideology
52
+ data_files:
53
+ - path: data/ideology/test.json
54
+ split: test
55
+ - path: data/ideology/dev.json
56
+ split: dev
57
+ - path: data/ideology/train.json
58
+ split: train
59
+ - config_name: conversations.molecular_entity
60
+ data_files:
61
+ - path: data/molecular_entity/test.json
62
+ split: test
63
+ - path: data/molecular_entity/dev.json
64
+ split: dev
65
+ - path: data/molecular_entity/train.json
66
+ split: train
67
  ---
68
+
69
+ # KGConv, a Conversational Corpus grounded in Wikidata
70
+
71
+ ## Table of Contents
72
+ - [Dataset Card Creation Guide](#dataset-card-creation-guide)
73
+ - [Table of Contents](#table-of-contents)
74
+ - [Dataset Description](#dataset-description)
75
+ - [Dataset Summary](#dataset-summary)
76
+ - [Languages](#languages)
77
+ - [Dataset Structure](#dataset-structure)
78
+ - [Data Instances](#data-instances)
79
+ - [Data Fields](#data-fields)
80
+ - [Additional Information](#additional-information)
81
+ - [Licensing Information](#licensing-information)
82
+ - [Citation Information](#citation-information)
83
+
84
+ ## Dataset Description
85
+
86
+ - **Repository:** [https://github.com/Orange-OpenSource/KGConv/]()
87
+ - **Paper:** [https://arxiv.org/abs/2308.15298]()
88
+ - **Point of Contact:** <quentin.brabant@orange.com>, <gwenole.lecorve@orange.com>, <linamaria.rojasbarahona@orange.com>, <claire.gardent@loria.fr>
89
+
90
+ ### Dataset Summary
91
+
92
+ KGConv is a large corpus of 71k english conversations where each question-answer pair is grounded in a Wikidata fact. The conversations were generated automatically: in particular, questions were created using a collection of 10,355 templates; subsequently, the naturalness of conversations was improved by inserting ellipses and coreference into questions, via both handcrafted rules and a generative rewriting model. The dataset thus provides several variants of each question (12 on average), organized into 3 levels of conversationality. KGConv can further be used for other generation and analysis tasks such as single-turn question generation from Wikidata triples, question rewriting, question answering from conversation or from knowledge graphs and quiz generation.
93
+
94
+
95
+ ### Languages
96
+
97
+ English.
98
+
99
+ ## Dataset Structure
100
+
101
+ ### Data Instances
102
+
103
+ Instance from the configs with name of the form "conversations.theme" (e.g. "conversations.country") have the following form:
104
+
105
+ ```
106
+ {
107
+ "conversation_id": "69795",
108
+ "root_neighbourhood": [
109
+ [
110
+ "Q6138903",
111
+ "P106",
112
+ "Q82955"
113
+ ],
114
+ [
115
+ "Q6138903",
116
+ "P19",
117
+ "Q3408680"
118
+ ],
119
+ ...
120
+ ],
121
+ "conversation": [
122
+ {
123
+ "triple": [
124
+ "Q691",
125
+ "P30",
126
+ "Q538"
127
+ ],
128
+ "question variants": [
129
+ {
130
+ "out-of-context": "In which continent is Papua New Guinea located?",
131
+ "in-context": "In which continent is Papua New Guinea located?",
132
+ "in-context subject ref": "Papua New Guinea",
133
+ "synthetic-in-context": "In which continent is Papua New Guinea located?"
134
+ },
135
+ {
136
+ "out-of-context": "In what continent is Papua New Guinea in?",
137
+ "in-context": "In what continent is Papua New Guinea in?",
138
+ "in-context subject ref": "Papua New Guinea",
139
+ "synthetic-in-context": "In what continent is Papua New Guinea in?"
140
+ },
141
+ ...
142
+ ],
143
+ "answer": "Oceania"
144
+ },
145
+ {
146
+ "triple": [
147
+ "Q691",
148
+ "P38",
149
+ "Q200759"
150
+ ],
151
+ "question variants": [
152
+ {
153
+ "out-of-context": "What is accepted as the currency of Papua New Guinea?",
154
+ "in-context": "What is accepted as the currency of Papua New Guinea?",
155
+ "in-context subject ref": "Papua New Guinea",
156
+ "synthetic-in-context": "What is accepted as the currency?"
157
+ },
158
+ {
159
+ "out-of-context": "What is the currency of Papua New Guinea?",
160
+ "in-context": "What is the currency of Papua New Guinea?",
161
+ "in-context subject ref": "Papua New Guinea",
162
+ "synthetic-in-context": "What is the currency?"
163
+ },
164
+ ...
165
+ ],
166
+ "answer": "kina"
167
+ },
168
+ ...
169
+ ```
170
+
171
+ Instances from the `labels` config are like this:
172
+
173
+ ```
174
+ {
175
+ "entity": "Q39",
176
+ "labels": [
177
+ "Swiss Confederation",
178
+ "CHE",
179
+ "Confoederatio Helvetica",
180
+ "Swiss",
181
+ "Schweiz",
182
+ "SUI",
183
+ "Switzerland",
184
+ "CH",
185
+ "Suisse",
186
+ "Svizzera"
187
+ ],
188
+ "preferred_label": "Switzerland"
189
+ }
190
+ ```
191
+
192
+ Instances from the `templates` config are as follows.
193
+
194
+ ```
195
+ {
196
+ "template_key": {
197
+ "p": "P1201",
198
+ "s_types": [
199
+ "Q149918"
200
+ ],
201
+ "o_types": []
202
+ },
203
+ "templates": [
204
+ {
205
+ "left": "what is the space tug of ",
206
+ "right": "?",
207
+ "source": "interface:automatic labeler"
208
+ },
209
+ {
210
+ "left": "what was the space tug of ",
211
+ "right": "?",
212
+ "source": "interface:624dc1cd4432b5035ba082df"
213
+ },
214
+ ...
215
+ ]
216
+ }
217
+ ```
218
+
219
+ ### Data Fields
220
+
221
+ The fields from the configs with name of the form "conversations.theme" (e.g. "conversations.country") are the following:
222
+
223
+ - `conversation`: list of dicts; each dict reprensent one question+answer and has the following fields:
224
+ - `conversation_id`: string
225
+ - `root_neighbourhood`: list of triples (each triple is itself represented by a list of 3 string elements) that constitute the neighbourhood of the conversation root entity in the knowledge graph (see the LREC publication for more details)
226
+ - `triple`: triple on which the question is based (list of three string elements)
227
+ - `question variants`: list of dict; each dict contain several forms of a question obtained via a given template (see the LREC publication for more details)
228
+ - `out-of-context`: one form of the question variant
229
+ - `in-context`: another form of the question variant
230
+ - `in-context subject ref`: how the subject is referred to in the in-context form
231
+ - `synthetic-in-context`: yet another form of the question variant
232
+ - `answer`: answer to the question (string)
233
+
234
+ The fields from the `labels` config are the following:
235
+
236
+ - `entity`: string, id of the entity
237
+ - `labels`: list of strings
238
+ - `preferred_label`: string
239
+
240
+ The fields from the `templates` config are the following:
241
+
242
+ - `template_key`: a dict containing the conditions for using the templates listed in `templates`, with the following fields:
243
+ - `p`: id of the property
244
+ - `s_types`: required types for subject
245
+ - `o_types`: require types for object
246
+ - `templates`: list of dicts representing templates; each dict has the following fields:
247
+ - `left`: left part of the template
248
+ - `right`: right part of the template
249
+ - `source`: origin of the template (string)
250
+
251
+
252
+ ## Additional Information
253
+
254
+ ### Licensing Information
255
+
256
+ This software is distributed under the Creative Commons Attribution 4.0 International,
257
+ the text of which is available at https://spdx.org/licenses/CC-BY-4.0.html
258
+ or see the "license.txt" file for more details.
259
+
260
+
261
+ ### Citation Information
262
+
263
+ ```
264
+ @article{brabant2023kgconv,
265
+ title={KGConv, a Conversational Corpus grounded in Wikidata},
266
+ author={Quentin Brabant and Gwenole Lecorve and Lina M. Rojas-Barahona and Claire Gardent},
267
+ year={2023},
268
+ eprint={2308.15298},
269
+ archivePrefix={arXiv},
270
+ primaryClass={cs.CL}
271
+ }
272
+ ```