HyperAIHyperAI
17 days ago

Benchmarking Geospatial Question Answering Engines using the Dataset GeoQuestions1089

{Pierre Maret, Manolis Koubarakis, Myrto Tsokanaridou, Michail Mitsios, Mariangela Pollali, Konstantinos Plas, Eleni Tsalapati, Dharmen Punjani, Sergios-Anestis Kefalidis}
Benchmarking Geospatial Question Answering Engines using the Dataset GeoQuestions1089
Abstract

We present the dataset GeoQuestions1089 for benchmarking geospatial question answering engines. GeoQuestions1089 is thelargest such dataset available presently and it contains 1089 questions, their corresponding GeoSPARQL or SPARQL queries and their answers over the geospatial knowledge graph YAGO2geo. We use GeoQuestions1089 to evaluate the effectiveness and efficiency of geospatial question answering engines GeoQA2 (an extension of GeoQA developed by our group) and the system of Hamzei et al. (2022).