Geo-Event Question Answering Systems: A Preliminary Research Study

Publication date

2021-09-01

Authors

Kazemi Beydokhti, Mohammad
Duckham, Matt
Griffin, Amy
Kasalica, VedranORCID 0000-0002-0097-1056ISNI 0000000492896480

Editors

Advisors

Supervisors

Document Type

Contribution to conference
Open Access logo

License

taverne

Abstract

Designing a Geospatial Question Answering (GeoQA) system that takes a user’s GIS-related domain question, understands how to gather the required data, how to analyse it, and how to present the results in a suitable format is arguably among the most important “moonshots” in the GeoAI field. In this study, we focus specifically on answering geo-event questions. This work begins by presenting a prototype process for generating workflows to answer geo-event questions by providing annotations of the domain, comprising a tool taxonomy we created from descriptions of geo-operations, a data type ontology obtained from the Core Concept Data types (CCD) ontology, and the annotations of the mentioned geo-operations with respect to the input/output pairs. Finally, the generated workflows are post-processed to restrict the solution space and provide more structured solutions. The results of this research provide a step towards the implementation of a geo-event QA system capable of answering diverse geo-event questions defined by users.

Keywords

Taverne

Citation

Kazemi Beydokhti, M, Duckham, M, Griffin, A & Kasalica, V 2021, 'Geo-Event Question Answering Systems: A Preliminary Research Study', pp. 1-6. https://doi.org/10.25436/E2KW2T