A Fuzzy Method to Measure Spatial Similarity in Geographic Information Retrieval

钟翔,刘家骏,高勇,李浩然

DOI: https://doi.org/10.3969/j.issn.1672-0504.2015.04.008

2015-01-01

Abstract:Geographical information retrieval is widely used to extract textually represented geographical knowledge in the Internet.When ranking the retrieval results,the similarity of spatial range should be taken into consideration.But the traditional calculation of spatial similarity ignores the fuzziness hidden in the textually represented geographical information,which leads to limitation and partiality.In this paper,we analysis deeply into textually represented geographical information,namely spatial assertion,and classify the fuzziness factors into reference-object fuzziness and spatial-relationship fuzziness.The uncertainty field is introduced to evaluate the influence degree of each factor.With the uncertainty field,we build the fuzzy footprints of documents and retrievals.On the basis of this,we propose a fuzzy method to measure spatial similarity in GIR,and rank the retrieval results according to their spatial similarities.Experiments show the method can reflect the fuzziness of spatial retrievals and measure the similarity between retrievals and documents better,and has a superior performance on ranking than traditional methods.