High-speed Detection of Ontological Knowledge and Bi-directional Lexico-Syntactic Patterns from the Web

摘要

We propose a high-speed method of detectingontological knowledge from the Web. Ontological knowledgein this paper means a term related to a given term. Forexample, hypernyms and hyponyms are basic related termsthat are treated in dictionaries. Synonyms and coordinateterms are also well-defined related terms. Topic terms anddescription terms represent topics of the given term and theyare vaguely defined. There are other related terms such asabbreviations and nicknames. The proposed method can beused for detecting many kinds of related terms. It extractsrelated terms from text resources only from Web searchresults, which consist of the titles, snippets, and URLs ofWeb pages. We use two different kinds of lexico-syntacticpatterns to extract related terms from the search results,and these are called bi-directional lexico-syntactic patterns.The proposed method can be applied to both languageswhere words are separated by a space such as English andKorean and ones where words are not separated by a spacesuch as Japanese and Chinese. The proposed method doesnot need any advanced natural language processing suchas morphological analysis or syntactic parsing. It worksrelatively fast and has excellent precision. We also propose amethod of automatically discovering superior bi-directionallexico-syntactic patterns using Web search engines becauseit is sometimes difficult to find appropriate patterns to detectrelated terms in a certain relationship.