Location Conflict Resolution with an Ontology

Location modelling is central for many pervasive applications and is a key challenge in this area. One major difficulty in location modelling is due to the nature of evidence about a person's location; the evidence is commonly noisy, uncertain and conflicting. Ontological reasoning is intuitively appealing to help address this problem, as reflected in several previous proposals for its use. This paper makes several important contributions to the exploration of the potential power of ontologies for improving reasoning about people's location from the available evidence. We describe ONCOR, our lightweight ontology framework: it has the notable and important property that it can be semi-automatically constructed, making new uses of it practical. This paper provides a comprehensive evaluation on how ontological reasoning can support location modelling: we introduce three algorithms for such reasoning and their evaluation based on a study of 8 people over 10---13 days. The results indicate the power of the approach, with mean error rates dropping from 55% with a naive algorithm to 16% with the best of the ontologically based algorithms. This work provides the first implementation of such an approach with a range of ontological reasoning approaches explored and evaluated.