WORD SENSE DISAMBIGUATION FOR AFAAN OROMO: USING KNOWLEDGE BASE

Abstract

Word sense disambiguation (WSD) is important and difficult problem that requires to be solved in Nature Language Processing. Afaan Oromo words have many meanings based on the context with which the word is used. In Afaan Oromo there is much ambiguous word in which there meaning is changing with the context. This creates the user of the language to be confused about the meaning of those words. In this paper we apply Knowledge based WSD method which is based on the database developed from scratch that uses Afaan Oromo Dictionary to disambiguate polysemous words in the sentence. The disambiguation process becomes accomplished based on words and sense relations developed in the database. The word sense disambiguation consists of preprocessing, morphological analysis, Afaan Oromo WordNet and disambiguation components to disambiguate ambiguous words of the language. Preprocessing component becomes the first stage to preprocess the input sentences to be used by morphological analysis to reduce the words to its root form or stem. The wordnet database stores words and it’s Synsets with their relation and concepts to disambiguate the polysemous words. Finally the disambiguation component disambiguates the ambiguous word using information from other components of word sense disambiguation that we use in this paper. Lastly, we conduct two experiments. The first experiment is with and without morphological analyzer that uses Afaan Oromo WordNet databases. The result of the experiment shows that an accuracy of 50.75% and 63.95% obtained. The second experiment becomes experiments that we conduct using various windows sizes to determine appropriate window sizes. According to the experiment window size of three- three becomes appropriate for Afaan Oromo.