Geographic information retrieval by topological, geographical, and conceptual matching

  • Authors:

  • Felix Mata

  • Affiliations:

  • PIIG Lab, Centre for Computing Research, National Polytechnic Institute, México, D.F., Mexico

  • Venue:
  • GeoS'07 Proceedings of the 2nd international conference on GeoSpatial semantics
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Geographic Information Science community is recognized that modern Geographic Information Retrieval systems should support the processing of imprecise data distributed over heterogeneous repositories. This means the search for relevant geographic results for a geographic query (QG) even if the data sources do not contain a result that matches exactly the user's request and then approximated results would be useful. Therefore, GIR systems should be centred at the nature and essence of spatial data (their relations and properties) taken into consideration the user's profile. Usually, semantic features are implicitly presented in different data sources. In this work, we use three heterogeneous data sources: vector data, geographic ontology, and geographic dictionaries. These repositories usually store topological relations, concepts, and descriptions of geographical objects under certain scenarios. In contrast to previous work, where these layers have been treated in an isolated way, their integration expects to be a better solution to capture the semantics of spatial objects. Thus, the use of spatial semantics and the integration of different information layers improve GIR, because adequate retrieval parameters according to the nature of spatial data, which emulate the user's requirements, can be established. In particular, we use topological relations {inside, in}, semantic relations {hyperonimy, meronimy}, and descriptions {constraints, representation}. An information extraction mechanism is designed for each data source, while the integration process is performed using the algorithm of ontology exploration. The ranking process is based on similarity measures, using the previously developed confusion theory. Finally, we present a case study to show some results of integrated GIR (iGIR) and compare them with Google's ones in a tabular form.