Automatic Acquisition of Noun Relations for Constructing Myanmar WordNet

Abstract

This paper describes the acquisition of noun relations for constructing Myanmar WordNet. WordNet is a useful lexical resource where specific senses of words are clustered together into synonymy sets, and semantic relationships between the sets are specified. WordNet is used in various NLP research, such as Information Extraction, Information Retrieval and in most other NLP application. The system has three steps. First, extract the lexico semantic relations by using LexicoSyntactic Pattern method. Second, by using information theoretic notion of mutual information, the new coming word has to be estimated to identify and in which the existing word of the association with sense. Third, refine the sense of noun word by manual. We have collected the noun word list (8943 words) and 55 patterns. We have obtained 87.9 % accuracy in sense identification. The system shows noun relationship between not only word level but also sense number. The system is implemented using Java.