Address Matching Based On Semantic Analysis Research

With the informatization and the construction of e-government in our country, the application of geographic space information resources and geographic information technology in various government departments such as urban planning, construction and management is increasingly wide. Currently, economic, cultural and social aspects of the data information of geographical feature space is not exist in geographic coordinates but in address text description way. It’s difficult to integrate, overlay and analyze with the superposition of spatial geographic data directly, limiting the information resource sharing, integration and comprehensive utilization. In fact the effect of many existing address matching methods is not ideal in practice. Especially for Chongqing, the mountain city, address matching met a bigger challenge because of address data with local features, problems left over by history and lack of strong regularity. Therefore, how to improve the success rate and accuracy of address matching has a very important practical significance for the space of information resources integration and sharing applications.This paper revolves around Chongqing address geocoding database construction and actual demand. After in-depth study of Chongqing address data present situation and characteristics, Chinese semantic address model construction method is put forward. This method improves the address model adaptability by optimizing the structure of existing address elements and setting up the spatial relationships between elements to increase the context binding. Based on Chinese semantic address model, this paper discusses standard address database design, database construction and quality control methods.Based on Chinese semantic address model and the standard library application, this paper joins the natural language processing methods in the research on the basis of traditional address matching algorithm. The technology route of address matching with semantic analysis is proposed. Introduction of support vector machine(SVM) in the address splitting can avoid ambiguity in Chinese natural language expression effectively for accuracy and efficiency improvement. To a certain extent, introduction of non-deterministic finite state machine to specification of address elements in the process of address resolution can solve the problem of the matching ambiguity of Chinese place names addresses, to improve the address matching success rate effectively. Using natural language processing methods in the address library matching to solve mismatch caused by wrong character and address ambiguity can improve address matching accuracy.Finally, taking an example of Nanan, compared with three traditional address matching method, through matching speed, matching rate and the matching accuracy, the address matching method based on semantic analysis is verified to have a better balance in comprehensive performance.