Study On Semantic Place Name And Address Matching Method Based On Geographic Information Public Service Platform

As a carrier to provide location services, geographic information public service platform brings benefits and convenience to urban construction, public life and economic development. A large number of place names and address data including in the thematic information of government, business and other functional departments, and natural language description of the place name and address strings using in the public location query process are all non-spatial information with semantic features. There is a need of techniques and methods to map them onto the map in geographic information public service platform in order to integrate and display, and then complete the visual analysis and other functions, so the establishment of effective linkages between semantic place name and address and spatial information will give the entire city with to a great help. Considering the problems of place name and address semantic matching and the needs for geographic information public service platform data integration, place name and address matching model in the combination form of typical administrative division name, street name, area name and house number is conducted in-depth study and research; Ontology model is used to represent geospatial entities and their interrelationships in natural language location description formally. Inference rules are defined to complete spatial relationship calculations, and thus natural language place name and address strings matching and spatial orientation are achieved. The validity and correctness of proposed method and model are validated through the algorithm implementation and experimental. New methods of semantic place name and address matching are researched and applied. The main contents are as follows:(1) Place name and address matching basic theory and key technology. Place name and address structure is researched in order to summarize the diverse description way and scalable model of address. Existing address matching methods are studied, and the model and algorithms of them are analyzed and compared, respectively. The advantages and shortcomings of different methods are given, and thus the address matching problems of existing methods are given. The problems are diversity and scalable address model matching and the natural language description of place name and address matching.(2) Multiple address element combination matching model construction and verification. Address records usually consist of multiple levels of address elements, including street, area and points of interest names. A multiple address element combination matching model is proposed. The demands of geographic information public service platform construction for address matching are introduced. The choice of address coding database as a reference data is analyses. The composition characteristic of address data of the city is studied, and on this basis the rule-based address segmentation method auxiliary by address word dictionaries is achieved. Combined strategy and matching process are studied, and the validation and evaluation of algorithm is conducted by matching experimental.(3) Geographic ontology and semantic matching inference mechanism. Because people do not have geographical information professional knowledge and have different cognitive structure, the description of geographic objects and locations are not all like administrative divisions, street names and other address elements. Therefore, more attention is needed on the characteristics of geographic relationship and semantics of geographic objects. Based on the analysis of human geospatial cognition feature, the theory and techniques of ontology and the geographic ontology are described, including basic concepts, building method, the spatial relationship representation and reasoning. The concept of semantic similarity is introduced to provide solution to the problem of address matching caused by cognition and natural language description.(4)Geographic ontology-based semantic matching model implementation and verification. A geographic ontology construction method is presented, while considering the problems of semantic address matching problem, geographic ontology model of urban areas is built and corresponding matching rules are defined. Based on this model natural language location description strings matching and positioning are completed, and through concrete examples to achieve verification of referential place name and address matching。...