Stemming Tigrinya Words for Information Retrieval
Omer Osman, Yoshiki Mikami · International Conference on Computational Linguistics · 2012
The increasing penetration of internet into less developed countries has resulted in the increase in the number of digital documents written in many minor languages. However, many of these languages have limited resources in terms of data, language resources and computational tools. Stemming is the reduction of inflected word forms into common basic form. It is an important analysis process in information retrieval and many natural language processing applications. In highly inflected languages such as Tigrinya, stemming is not always straightforward task. In this paper we present the development of stemmer for Tigrinya words to facilitate the information retrieval. We used a hybrid approach for stemming that combines rule based stemming which removes affixes in successively applied steps and dictionary based stemming which reduces stemming errors by verifying the resulting stem based on word distance measures. The stemmer was evaluated using two sets of Tigrinya words. The results show that it achieved an average accuracy of 89.3%.