Corpus-based Ontology Learning for Word Sense Disambiguation

Corpus-based Ontology Learning for Word Sense Disambiguation Sin-Jae Kang School of Computer and Communication Engineering, Daegu University and Advanced Information Technology Research Center (AITrc) Nairi-Ri 15, Jinrang, Kyungsan, Kyungpook, 712-714 Republic of KOREA sjkang@daegu.ac.kr Abstract This paper proposes to disambiguate word senses by corpus-based ontology learning. Our approach is a hybrid method. First, we apply the previously-secured dictionary information to select the correct senses of some ambiguous words with high precision, and then use the ontology to disambiguate the remaining ambiguous words. The mutual information between concepts in the ontology was calculated before using the ontology as knowledge for disambiguating word senses. If mutual information is regarded as a weight between ontology concepts, the ontology can be treated as a graph with weighted edges, and then we locate the least weighted path from one concept to the other concept. In our practical machine translation system, our word sense disambiguation method achieved a 9% improvement over methods which do not use ontology for Korean translation. 1. Introduction An ontology is a knowledge base with information about concepts existing in the world, their properties, and how they relate to each other. An ontology is different from a thesaurus in that it contains only language independent information and many other semantic relations, as well as taxonomic relations. In this paper, we propose to use the ontology to disambiguate word senses. All approaches to word sense disambiguation (WSD) make use of words in a sentence to mutually disambiguate each other. The distinctions between various approaches lie in the source and type of knowledge made by the lexical units in a sentence. Thus, all these approaches can be classified into Al-based, knowledge-based, or corpus-based approaches, according to their sources and types of knowledge (Ide, 1998). AI-based WSD methods (Dahlgren, 1988) use a semantic network, or frames containing information about word functions and the relation to other words in individual sentences; or preference semantics, which specifies selectional restrictions for combinations of lexical items in a sentence. The difficulty with handcrafting the knowledge sources is the major disadvantage of AI-based systems. Knowledge-based methods (Resnik, 1995a; Yarowsky, 1992) have utilized machine-readable dictionaries (MRD), thesauri, and computational lexicons, such as WordNet. Since most MRDs and thesauri were created for human use and display inconsistencies, these methods have clear limitations. Corpus-based methods (Dagan, 1994; Gale, 1992) extract statistical information from corpora which is monolingual or bilingual, and raw or sense-tagged. The problem of data sparseness commonly occurs in the corpus-based approach, and is especially severe when processing in WSD. A smoothing and concept-based method is used to address this problem. Our WSD approach is a hybrid method, which combines the advantages of corpus-based and knowledge-based methods. We use our semi-automatically constructed ontology as an external knowledge source and secured dictionary information as context information. First, we apply the previously-secured dictionary information to select the correct senses of some ambiguous words with high precision, and then use the ontology to disambiguate the remaining ambiguous words. The remainder of this paper is …

Content truncated to 500 words