Mining type information from Chinese online encyclopedias

Abstract

Recently, there is an increasing interest in extracting or mining type information from Web sources. Type information stating that an instance is of a certain type is an important component of knowledge bases. Although there has been some work on obtaining type information, most of current techniques are either language-dependent or to generate one or more general types for a given instance because of type sparseness. In this paper, we present a novel approach for mining type information from Chinese online encyclopedias. More precisely, we mine type information from abstracts, infoboxes and categories of article pages in Chinese encyclopedia Web sites. In particular, most of the generated Chinese type information is inferred from categories of article pages through an attribute propagation algorithm and a graph-based random walk method. We conduct experiments over Chinese encyclopedia Web sites: Baidu Baike, Hudong Baike and Chinese Wikipedia. Experimental results show that our approach can generate large scale and high-quality Chinese type information with types of appropriate granularity.