Neural factoid geospatial question answering

Abstract

Existing question answering systems struggle to answer factoid questions when geospatial information is involved. This is because most systems cannot accurately detectthe geospatial semantic elements from the natural language questions, or capture thesemantic relationships between those elements. In this paper, we propose a geospatial semanticencoding schema and a semantic graph representation which captures the semanticrelations and dependencies in geospatial questions. We demonstrate that our proposedgraph representation approach aids in the translation from natural language to a formal,executable expression in a query language. To decrease the need for people to provideexplanatory information as part of their question and make the translation fully automatic,we treat the semantic encoding of the question as a sequential tagging task, and the graphgeneration of the query as a semantic dependency parsing task. We apply neural networkapproaches to automatically encode the geospatial questions into spatial semantic graphrepresentations. Compared with current template-based approaches, our method generalisesto a broader range of questions, including those with complex syntax and semantics.Our proposed approach achieves better results on GeoData201 than existing methods