Word Sense Disambiguation of Farsi Homographs Using Thesaurus and Corpus

摘要

This paper describes disambiguation of Farsi homographs in unrestricted text using thesaurus and corpus. The proposed method is based on with some differences. These differences consist of first using collocational information to avoid the collection of spurious contexts caused by polysemous words in thesaurus categories, and second contribution of all words in the test data context, even those not appeared in the collected contexts to the calculation of the conceptual classes' score. Using a Farsi corpus and a Farsi thesaurus, this method correctly disambiguated 91.46% of the instances of 15 Farsi homographs. This method was compared to three supervised corpus based methods including Naieve Bayes, Exemplar-based, and Decision List. Unlike supervised methods, this method needs no training data, and has a good performance on disambiguation of uncommon words. In addition, this method can be used for removing some kinds of morphological ambiguities.