Word Sense Disambiguation by Relative Selection

Word Sense Disambiguation by Relative Selection Hee-Cheol Seo1, Hae-Chang Rim2, and Myung-Gil Jang1 1 Knowledge Mining Research Team, Electronics and Telecommunications Research Institute (ETRI), Daejeon, Korea {hcseo, mgjang}@etri.re.kr 2 Dept. of Computer Science and Engineering, Korea University, 1, 5-ka, Anam-dong, Seongbuk-Gu, Seoul, 136-701, Korea rim@nlp.korea.ac.kr Abstract. This paper describes a novel method for a word sense disam- biguation that utilizes relatives (i.e. synonyms, hypernyms, meronyms, etc in WordNet) of a target word and raw corpora. The method disam- biguates senses of a target word by selecting a relative that most prob- ably occurs in a new sentence including the target word. Only one co- occurrence frequency matrix is utilized to efficiently disambiguate senses of many target words. Experiments on several English datum present that our proposed method achieves a good performance. 1 Introduction With its importance, a word sense disambiguation (WSD) has been known as a very important field of a natural language processing (NLP) and has been studied steadily since the advent of NLP in the 1950s. In spite of the long study, few WSD systems are used for practical NLP applications unlike part-of-speech (POS) taggers and syntactic parsers. The reason is because most of WSD studies have focused on only a small number of ambiguous words based on sense tagged corpus. In other words, the previous WSD systems disambiguate senses of just a few words, and hence are not helpful for other NLP applications because of its low coverage. Why have the studies about WSD stayed on the small number of ambiguous words? The answer is on sense tagged corpus where a few words are assigned to correct senses. Since the construction of the sense tagged corpus needs a great amount of times and cost, most of current sense tagged corpora contain a small number of words less than 100 and the corresponding senses to the words. The corpora, which have sense information of all words, have been built recently, but are not large enough to provide sufficient disambiguation information of the all words. Therefore, the methods based on the sense tagged corpora have difficulties in disambiguating senses of all words. R. Dale et al. (Eds.): IJCNLP 2005, LNAI 3651, pp. 920–932, 2005. c ⃝ Springer-Verlag Berlin Heidelberg 2005 Word Sense Disambiguation by Relative Selection 921 In this paper, we proposed a novel WSD method that requires no sense tagged corpus1 and that identifies senses of all words in sentences or documents, not a small number of words. Our proposed method depends on raw corpus, which is relatively very large, and on WordNet [1], which is a lexical database in a hierarchical structure. 2 Related Works There are several works for WSD that do not depend on a sense tagged corpus, and they can be classified into three approaches according to main resources used: raw corpus based approach [2], dictionary based approach [3,4] and hier- archical lexical database approach. The hierarchical lexical database approach can be reclassified into three groups according to usages of the database: gloss based …

Content truncated to 500 words