Study Of Chinese POI Name Matching Based On Role Tagging And Prototype System

Address Geocoding is a key technologies for connecting isolated information systems and integrating information resources in Digital City. POI name matching is one of the key technologies for Address Geocoding. This paper analyzes the shortcoming of existed Chinese POI name macthing methods. We present a role tagging-based way using Chinese word segmentaion,information retrieval (IR),search engine (SE) and artificial Intelligence technologies , which can improve the rate of POI name macthing success. That is: tokens after segmentation are tagged using HMM with diffirent roles according to theirs functions in the generation of Chinese POI s. Tokens with diffirent roles have diffirent degree of importance in matching. After exact matching failure. Fuzzy matching is excuted according to tokens' roles.A prototype system is implemented to test the actual performance of role tagging-based way . A matching experiment was carried out, which uses 2300 POIs chosen randomly from more than 160000 POIs in Dongcheng District and Xicheng District of Beijing city. The rate of matching achieved success 92. 39%. Detailed analysis using about 2300 sample results showed that the main problems limiting the matching success rate are the quality of the POI database and limited fuzzy matching capability of algorithm. Excluding the effect of the POI database, the method itself achieved 98.01% matching success rate. Enhancement ideas and possible works and emphasis are given at the end of the thesis.