GeoChanges QA
Authors/Creators
- 1. Dept. of Informatics and Telecommunications, National and Kapodistrian University of Athens
- 2. L3S Research Center, Leibniz Universit ̈at Hannover, Germany
- 3. Data Science & Intelligent Systems Group (DSIS), University of Bonn and Lamarr Institute for Machine Learning and Artificial Intelligence, Bonn, Germany
- 4. Dept. of Informatics and Telecommunications, National and Kapodistrian University of Athens, Greece
Description
This work introduces GeoChangesQA, a novel spatiotemporal QA dataset for historical geospatial knowledge. We first developed the Historical County Boundaries Ontology (HCB-O) to create this dataset. Then, we leveraged geospatial historical data spanning regions in the United States from 1630 to 2000 to populate the novel Historical County
Boundaries Knowledge Graph (HCB-KG). Subsequently, we developed a semi-automated procedure for generating questions, GeoSPARQL queries, and corresponding answers over HCB-KG by leveraging subgraph and query template extraction techniques. Through this method, we automatically generated 5, 700 question-query-answer triples.
Series information (English)
Columns Description:
- Q_id: Unique Subgraph type id
- Instances: Instances that may be used inside SPARQL Query
- Select_type: The target node of the query, what the question asks for
- Predicates: The predicates used in the query
- Filters: The Filters that may be used in the query. Has the format of {type of the filter, variable1, operator, variable2}
- SparqlQuery: The SPARQL query
- Uris_match: The matching between question template variables and their values
- Questions_templates: The selected Question template for the query
- Questions: The Question after replacing Uris_match with template variables
- Corrected Questions: The question after passing through the grammar corrector model. If nan no change was needed.
- Answers: One sample answer to the SPARQL query
Files
GeoChanges_QA_Dataset.csv
Files
(8.7 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:47c48991b20e52b003b4fe93247b99c6
|
8.7 MB | Preview Download |