DESIGNING AND DEVELOPING A SPELL CHECKER MODEL FOR WOLAITA LANGUAGE USING DEEP LEARNING APPROACH

Abstract

Natural language processing (NLP) is one of the subfield of artificial intelligence (AI) that empowers the interaction between human and computer to get it and translate common language as people do. In a NLP, there are different techniques are applied to identify and extract the natural language rules. Particularly, a spell checker is one of the techniques that helps the human being to check the validity of words in the large collection of documents. Wolaita language is used by different government and non-government organizations for various purposes like school teaching and learning processes, for news in the Wolaita TV, administration language in Wolaita zone, as a field of study in the Country University, that produces large documents and during these documents preparation there are serious spelling problem in the processes. However, the absence of a spell checker model for Wolaita language has made document preparation activities difficult and challenging that needs excessive effort to edit and correct documents, reduce documents quality and time wastage. Thus, these spelling errors can occur when people use text processing application to produce electronic documents. And to the knowledge of this researcher, no spell checker systems have been developed for the Wolaita language. Therefore, in this study, the aim of the researcher is to design and develop a spell checker model for Wolaita language on variety of Wolaita language text words. In doing this, the study was being applied text character level spell checker techniques that was designed based on encoder-decoder architecture. Moreover, the study was being applied different deep learning algorithms such as BiLSTM, LSTM, GRU and RNN with suitable feature extraction techniques. Finally, the experimental results of selected three models are 98.53%, 95.97%, 91.35% and 86.37% respectively.