Hybrid Distance Based Measures for Geospatial Domain

Current researchers of search engines focus more on semantic based information retrieval as syntax based retrieval yield less precision. Retrieving relevant information from diverse heterogeneous web resources remains as a challenge. Distance based measures play a major role in the information retrieval systems. This work focuses on retrieving relevant concepts using geospatial datasources to aid geospatial applications in predicting floods, locating underground pipes and cables and testing the quality of water. Geospatial data characterizes geographical features of the real world using spatial extent and location. This paper proposes HDSM algorithm for geospatial information retrieval which adapts the existing distance based measures viz., Manhattan distance, Euclidean distance, Vector cosine and bray Curtis for the geospatial domain to identify related concepts to the geo-spatial query concept. All these four proposed hybrid distance based measures combine the advantages of geometric and network semantic similarity models. The meaning of the geospatial concepts are captured from the expressive knowledge of the geospatial concept properties and geospatial relations. These proposed four Hybrid distance based measures have been tested using Ordnance Survey Master Map data source and ordnance survey ontology for varying semantic similarity thresholds. The experimental results are reported in this paper. The Hybrid Manhattan distance based measure has yielded the precision of 95%.