Discovering ISO relations among synsets in a flat thesaurus automatically

The semantic resources are very important for natural language processing (NLP). Constructing the semantic resources by hand can lead to good results, but costs a lot. This paper proposes an effective method to automatically construct a hierarchically organized resource based on a traditional flat thesaurus. With the method, we could enrich immediate super-ordinate (ISO) relations among synsets in the existing resources and make good use of them in the field of NLP. We also develop an automatic method to evaluate the semantic relations detected in our experiment with a machine-readable dictionary.