Heuristic to Build RCC8 for Event Locations
Majed Ayyad
DOI: http://dx.doi.org/10.15439/2014F353
Citation: Proceedings of the 2014 Federated Conference on Computer Science and Information Systems, M. Ganzha, L. Maciaszek, M. Paprzycki (eds). ACSIS, Vol. 2, pages 299–306 (2014)
Abstract. Events that are detected and reported by humans to actionable knowledge bases in multi-tier responding agencies have significant amount of spatial information. Humans have intuitive ability to triage repeated or duplicated events based on their spatio-temporal information. However, this cognitive process is not modeled easily and human ability is limited in situations where large number of events are reported simultaneously. The likelihood of two events to be the same is higher if they occur on the same place and time. In this work, we focus only on calculating location equivalence of events. For this purpose we use RCC8 theory to represent spatial relations between regional locations. The algorithm designed approximates the arbitrary shape of regions into circles and build region connection relations based on the size of the circle. The end result is a region of circular tiles with explicit RCC8 relations that could be used to reason on the relation between the locations of events. Additionally, we outline some experiments to evaluate the precision and recall of the results based on the used corpus. These results indicate that although the task is challenging, automated methods are capable of building spatial regional relations between events