Automatic construction and visualization of address models

Addresses are one of the most popular geographical reference systems in natural language. Addresses are considered as the most effective approach to bridging the gap between business data in management information system (MIS) and GIS, which supports geospatial information visualization and spatial analysis. One significant problem in Chinese address geocoding is the irregular address expressions, which makes it impossible to get a nationwide address content standardization. In order to solve this problem, this paper presents an approach to automatically extraction and visualization of address models. In this approach, we focus on the binary relationship between two address elements rather than address templates. Based on the classification of address elements and splitting of large-scale Chinese addresses, we use boolean association rules to induce address models. The experimental results indicate that the proposed approach is more flexible and practical in analyzing Chinese address structures for geocoding service.