Automatic Extending HowNet's Attribute Lexicon on the Web

Abstract

It is well known that lexical knowledge sources as WordNet, HowNet are very important to natural language processing applications in artificial intelligence. This paper proposes a new Web-mining algorithm to automatically extend HowNetpsilas attribute set. Seeded by the original attribute set in HowNet, the algorithm uses a conjunctive lexical pattern plus a validation mechanism called PES (position exchanging search) to extend the lexicon iteratively on the Web. Also, a Web-based attribute classifier is constructed which behaves as a filter to control the level of false positives during each iteration. The algorithm is evaluated using both standard human judgements and HowNet based evaluation. Some experimental results about the performance of the method are provided.