Large Scale Qualitative Spatio-Temporal Reasoning

This thesis considers qualitative spatio-temporal reasoning (QSTR), a branch of artificial intelligence that is concerned with qualitative spatial and temporal relations between entities. Despite QSTR being an active area of research for many years, there has been comparatively little work looking at large scale qualitative spatio-temporal reasoning - reasoning using hundreds of thousands or millions of relations. The big data phenomenon of recent years means there is now a requirement for QSTR implementations that will scale effectively and reason using large scale datasets. However, existing reasoners are limited in their scalability, what is needed are new approaches to QSTR.

This thesis considers whether parallel distributed programming techniques can be used to address the challenges of large scale QSTR. Specifically, this thesis presents the first in-depth investigation of adapting QSTR techniques to work in a distributed environment. This has resulted in a large scale qualitative spatial reasoner, ParQR, which has been evaluated by comparing it with existing reasoners and alternative approaches to large scale QSTR. ParQR has been shown to outperform existing solutions, reasoning using far larger datasets than previously possible.

The thesis then considers a specific application of large scale QSTR, querying knowledge graphs. This has two parts to it. First, integrating large scale complex spatial datasets to generate an enhanced knowledge graph that can support qualitative spatial reasoning, and secondly, adapting parallel, distributed QSTR techniques to implement a query answering system for spatial knowledge graphs. The query engine that has been developed is able to provide solutions to a variety of spatial queries. It has been evaluated and shown to provide more comprehensive query results in comparison to using quantitative only techniques.