Question Answering Engines for Geospatial Knowledge Graphs

Abstract

Question answering (QA) has been a hot topic of research in Natural Language Processing [Rajpurkar et al. 2016] and Knowledge Graphs for a number of years [Höffner et al. 2017, Diefenbach et al. 2018, Roy and Anand 2021]. There has been a lot of work on the construction of training datasets, the design of effective QA algo­ rithms, and the benchmarking of implemented systems. Very recently, work has concentrated on QA systems integrating ideas from Language Models, Knowledge Graphs, and Neural Networks [Yasunaga et al. 2021]. Question answering techniques have also been adopted by industry. They are used in all well-known search engines (e.g., Google and Bing), digital assistants (e.g., Siri, Alexa, and Google Assistant), and chatbots such as BlenderBot 2.0 of Facebook.1 The need for geospatial question answering engines arises in many practical situ­ ations. For example, a tourist visiting Athens would like to know which museums have parks nearby so that she can rest in the shade after visiting the museum. Or when driving toward Ancient Olympia, a tourist might want to know which Greek restaurant along the way is more popular with locals. Finally, a question answering engine might drive the implementation of a geography tutoring system by being able to answer questions such as “What is the tallest mountain in Greece?” or “What countries border Greece and are these countries members of the European Union?” Question Answering Engines for Geospatial Knowledge Graphs