Paradigm Model based Hybrid Morph Analyzer for Gujarati using Partial Stemmer

Morphological Analysis plays a major role in understanding and generating words and their forms for an efficient Machine Translation system. Morph Analyzer analyzes a word form, provides us the root or stem word and the linguistic information attached to the word form. It also generates a word form based on the root word and the linguistic information as provided. This linguistic information is very helpful during every stages of Natural Language Processing. In this paper we propose a Morph Analyzer which we have built keeping Gujarati language in mind. A Morph Analyzer can be made more efficient if the target language is known. We have added a partial stemmer in the paradigm model of a Morph Analyzer to achieve better accuracy. Our approach is the supervised approach where we manually build a dictionary of root words and classify them into different paradigms. A built-in partial stemmer performs the analysis and generation of word forms based on language dependent linguistic rules. Using this approach on Gujarati language we were able to achieve a considerable coverage of approximately 5,000 nouns.