A Practical Approach to Extracting Names of Geographical Entities and Their Relations from the Web

摘要

Geographical information extraction is a special case of information extraction. In this paper, we present a practical method of extracting both names of geographical entities and their relations from the Web. The method is composed of three major phases. First, we manually designed a list of 493 Chinese lexico-syntactical patterns for matching Web page excerpts which contain names of geographical entities and their relations; second, we developed a knowledge extractor for extracting those names and relations to generate a geographical graph whose nodes are entities, and edges represent relations of the entities; third, we developed several methods for handling problems or errors in the generated graph. Experimental results show that the OMKast-Googling system has a satisfactory performance both in the entity name extraction and relation extraction.