NLPRL Odia-English: Indic Language Neural Machine Translation System

Abstract

In this paper, we (team name is NLPRL) describe our systems that were submitted to the translation shared tasks at WAT 2020. We submitted two systems Odia to English and English to Odia translation under Indic tasks. We presented the Neural Machine Translation (NMT) system based on the Transformer approach using a byte-level version of BPE, which requires a base vocabulary of size 256 only. Experiments show the BLEU score of English to Odia 1.34 and Odia to English 11.33 on the benchmark test data. ©2020 Association for Computational Linguistics