Transformer-based neural machine translation system for Tamil - English

Abstract

This paper describes the Machine Translation (MT) system submitted by the NLPRL team for the Tamil - English Indic Task at WAT 2019. We presented the Neural Machine Translation (NMT) system based on the Transformer approach. Training and performance of the model are evaluated on the EnTam corpus (An English-Tamil Parallel Corpus) collected by researchers at UFAL (Institute of Formal and Applied Linguistics). The evaluation of the model done using Adequacy, BLEU, RIBES, and AM-FM scores, and the model improves translation in terms of Adequacy, RIBES and AM-FM as compared to the baseline. © 2019 Association for Computational Linguistics