Integration of heterogeneous language resources: A monolingual dictionary and a thesaurus

Linguistic knowledge plays a crucial role in natural language processing. Constructing large linguistic knowledge bases requires a lot of human effort and much cost. There have been many attempts to construct linguistic knowledge automatically, based on two primary strategies: knowledge extraction from annotated corpora and the augmentation of existing knowledge bases using annotated corpora. This paper describes an algorithm to enlarge existing linguistic knowledge through integration with heterogeneous linguistic resources. Specifically, this algorithm links a word sense defined in a monolingual dictionary to semantic classes in a thesaurus. Experiments show that we achieve a linking precision of 85.5% and coverage of 61.4%.