A Novel Kernel Regression Based Machine Translation System for Sinhala-Tamil Translation

Abstract

Sinhala-Tamil Machine Translation is becoming one of the active researches in Natural Language Processing eld. This is a challenging and interesting scientific problem. The language pair has shown structural similarity and has produced a relatively good result in past researches. This thesis investigates the applicability of Kernel Ridge Regression technique to Sinhala-Tamil translation with small Data set. For this, the translation problem is viewed as a string-to-string mapping among source and target language features. Detailed description about all key components of our system, including the Kernel Ridge Regression model, two novel solutions for pre-image problem and the decoder are presented. Interestingly, with the evaluation results we nd that we can learn the linear correspondence in Sinhala and Tamil languages and also with that we can do the translation.