Datasets:
Enriched readme
Browse files
README.md
CHANGED
@@ -102,3 +102,113 @@ dataset_info:
|
|
102 |
download_size: 9637685
|
103 |
dataset_size: 42808841
|
104 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
102 |
download_size: 9637685
|
103 |
dataset_size: 42808841
|
104 |
---
|
105 |
+
|
106 |
+
# Dataset Card for WEBNLG-QA
|
107 |
+
|
108 |
+
## Dataset Description
|
109 |
+
|
110 |
+
- **Paper:** [SPARQL-to-Text Question Generation for Knowledge-Based Conversational Applications (AACL-IJCNLP 2022)](https://aclanthology.org/2022.aacl-main.11/)
|
111 |
+
- **Point of Contact:** Gwénolé Lecorvé
|
112 |
+
|
113 |
+
### Dataset Summary
|
114 |
+
|
115 |
+
WEBNLG-QA is a conversational question answering dataset grounded on WEBNLG. It consists in a set of question-answering dialogues (follow-up question-answer pairs) based on short paragraphs of text. Each paragraph is associated a knowledge graph (from WEBNLG). The questions are associated with SPARQL queries.
|
116 |
+
|
117 |
+
### Supported tasks
|
118 |
+
|
119 |
+
The dataset can be used for:
|
120 |
+
* Knowledge-based question-answering
|
121 |
+
* SPARQL-to-Text conversion
|
122 |
+
|
123 |
+
|
124 |
+
#### Knowledge based question-answering
|
125 |
+
|
126 |
+
Below is an example of dialogue:
|
127 |
+
- Q1: What is used as an instrument is Sludge Metal or in Post-metal?
|
128 |
+
- A1: Singing, Synthesizer
|
129 |
+
- Q2: And what about Sludge Metal in particular?
|
130 |
+
- A2: Singing
|
131 |
+
- Q3: Does the Year of No Light album Nord belong to this genre?
|
132 |
+
- A3: Yes.
|
133 |
+
|
134 |
+
#### SPARQL-to-Text Question Generation
|
135 |
+
|
136 |
+
SPARQL-to-Text question generation refers to the task of converting a SPARQL query into a natural language question, eg:
|
137 |
+
|
138 |
+
```SQL
|
139 |
+
SELECT (COUNT(?country) as ?answer)
|
140 |
+
WHERE { ?country property:member_of resource:Europe .
|
141 |
+
?country property:population ?n .
|
142 |
+
FILTER ( ?n > 10000000 )
|
143 |
+
}
|
144 |
+
```
|
145 |
+
|
146 |
+
could be converted into:
|
147 |
+
|
148 |
+
```txt
|
149 |
+
How many European countries have more than 10 million inhabitants?
|
150 |
+
```
|
151 |
+
|
152 |
+
## Dataset Structure
|
153 |
+
|
154 |
+
### Types of questions
|
155 |
+
|
156 |
+
Questions are very diverses, ranging the following feature categories:
|
157 |
+
* SPARQL query
|
158 |
+
* Number of triplets: 1, 2, 3+
|
159 |
+
* Logical connector between clauses: conjunction, disjunction, exclusion
|
160 |
+
* Topology of the query graph: direct, sibling, chain, mixed, other
|
161 |
+
* Variable typing: none, target variable, internal variable
|
162 |
+
* Comparison clause: none, string, number, date
|
163 |
+
* Answer
|
164 |
+
* Type: entity (open), entity (close list of choices), number, boolean
|
165 |
+
* Number of answers: 0 (unanswerable question), 1, 2+
|
166 |
+
* Number of target variables: 0 ("ASK" verb), 1, 2
|
167 |
+
* Contextuality
|
168 |
+
* Dialogue context: self-sufficient question, contains coreference, contains ellipsis
|
169 |
+
* Meaning: meaninful question, non-sense question
|
170 |
+
|
171 |
+
|
172 |
+
### Data splits
|
173 |
+
|
174 |
+
Text verbalization is only available for a subset of the test set, referred to as *challenge set*. Other sample only contain dialogues in the form of follow-up sparql queries.
|
175 |
+
|
176 |
+
| | Train | Validation | Test | Challenge |
|
177 |
+
| --------------------- | ---------- | ---------- | ---------- | ------------ |
|
178 |
+
| Questions | 27727 | 3485 | 4179 | 332 |
|
179 |
+
| Dialogues | 1001 | 1264 | 1417 | 100 |
|
180 |
+
| NL question per query | 0 | 0 | 0 | 2 |
|
181 |
+
| Characters per query | 129 (± 43) | 131 (± 45) | 122 (± 45) | 113 (± 38) |
|
182 |
+
| Tokens per question | - | - | - | 8.4 (± 4.5) |
|
183 |
+
|
184 |
+
|
185 |
+
## Additional information
|
186 |
+
|
187 |
+
### Related datasets
|
188 |
+
|
189 |
+
This corpus is part of a set of 5 datasets released for SPARQL-to-Text generation, namely:
|
190 |
+
- Non conversational datasets
|
191 |
+
- [SimpleQuestions](https://huggingface.co/datasets/OrangeInnov/simplequestions-sparqltotext) (from https://github.com/askplatypus/wikidata-simplequestions)
|
192 |
+
- [ParaQA](https://huggingface.co/datasets/OrangeInnov/paraqa-sparqltotext) (from https://github.com/barshana-banerjee/ParaQA)
|
193 |
+
- [LC-QuAD 2.0](https://huggingface.co/datasets/OrangeInnov/lcquad_2.0-sparqltotext) (from http://lc-quad.sda.tech/)
|
194 |
+
- Conversational datasets
|
195 |
+
- [CSQA](https://huggingface.co/datasets/OrangeInnov/csqa-sparqltotext) (from https://amritasaha1812.github.io/CSQA/)
|
196 |
+
- [WebNLQ-QA](https://huggingface.co/datasets/OrangeInnov/webnlg-qa) (derived from https://gitlab.com/shimorina/webnlg-dataset/-/tree/master/release_v3.0)
|
197 |
+
|
198 |
+
### Licencing information
|
199 |
+
|
200 |
+
* Content from original dataset: CC-BY-SA 4.0
|
201 |
+
* New content: CC BY-SA 4.0
|
202 |
+
|
203 |
+
|
204 |
+
### Citation information
|
205 |
+
|
206 |
+
```bibtex
|
207 |
+
@inproceedings{lecorve2022sparql2text,
|
208 |
+
title={SPARQL-to-Text Question Generation for Knowledge-Based Conversational Applications},
|
209 |
+
author={Lecorv\'e, Gw\'enol\'e and Veyret, Morgan and Brabant, Quentin and Rojas-Barahona, Lina M.},
|
210 |
+
journal={Proceedings of the Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the International Joint Conference on Natural Language Processing (AACL-IJCNLP)},
|
211 |
+
year={2022}
|
212 |
+
}
|
213 |
+
```
|
214 |
+
|