Deep Learning Approach for Sidaamu Afoo to English Machine Translation

Abstract

Advances in computer applications and systems provide solutions to our everyday problems. Machine learning is one of the computational applications of algorithms and statistical mod els that uses algorithms and statistical models to perform tasks without explicit instructions, but uses models for reasoning. Recently, neural machine translation (NMT) has been im proved by selectively focusing on specific parts of the source sentence during translation us ing an attentional mechanism. The study of practical architectures for attention-based NMT hasn't received much attention, though. The researcher motivated by that of no previous at tempts have been made as per the In this research an attempt have been made to experiment on Sidaama to English Machine Translation using Deep Learning using an attention mecha nism for promoting information sharing. Since there is no English and Sidaama parallel text corpus, the researchers collected 24467 a parallel text corpus for Sidaama to English Ma chine Translation system from religious domain specifically from bible, health questionnaire, and other online available data. The researchers conducted series of experiments on various models namely Statistical Machine Translation (SMT), Recurrent Neural Network (RNN) with encoder-decoder attention model and RNN with Long Short Term Memory (LSTM) ap proaches and achieved an average accuracy result of 72%, 83.3% and 83.2% respectively. The researcher is now working on post-editing to enhance the performance of the bi-lingual English-Sidaama translator.