Although many algorithms have been developed to harvest lexical knowledge from text, few of them mined terms into taxonomies. In this paper we present an approach aimed at learning a lexical taxonomy automatically starting from an unstructured corpus. Previous algorithms for taxonomy induction have typically focused on discovering new single relationships based on hand-constructed or automatically discovered textual patterns or clustering method. In this paper we use the advantages of pattern base method and new semantic relationships are extracted from web pages. Strength of the present method is word sense disambiguation from extracted relations. Our experiments show that we obtain high-quality results. Our system achieves higher F-measure than other methods, and its flexible design of the framework allows to automatically reconstruct WordNet-like taxonomies.
