Abstract: With the development of spontaneous geographic information and Chinese address element segmentation technology, the quality of address elements needs to be evaluated. Aiming at the problem that the quality of address elements produced by Chinese address text segmentation is difficult to effectively evaluate, this paper proposes a method for evaluating the credibility of address elements supported by multi-source data and network retrieval. Firstly, the Chinese word segmentation tool is used to segment the address elements and part-of-speech tagging. By analyzing the word frequency and part-of-speech combination mode, the credibility of the naming structure of the address elements is calculated. Then, based on large-scale address samples, road data, and POI data, excavate the data support of multi-source data to address elements, and calculate the data support. Then use the search engine to retrieve the address elements quickly, analyze the search results and quantity, and calculate the network credibility of the address elements. Finally, a comprehensive credibility calculation model for address elements is proposed to realize the comprehensive credibility calculation of address elements. Experimental results show that the model and method can not only efficiently and quickly calculate the credibility of address elements in Chinese address texts, but also effectively discover the remoteness and falsehood of address elements, which provides a reference for the automatic detection and standardization of address elements.

Key words: multi-source data, credibility evaluation, Chinese word segmentation, address element, information normalize