Neural Machine Translation Approach for Singlish to English Translation

Abstract

Comprehension of “Singlish” (an alternative writing system for Sinhala language) texts by a machine had been a requirement for a long period. It has been a choice of many Sri Lankan’s writing style in casual conversations such as small talks, chats and social media comments. Finding a method to translate Singlish to Sinhala or English has been tried for a couple of years by the research community in Sri Lanka and many of the attempts were tried based on statistical language translation approaches due to the challenge of finding a large dataset to use Deep Learning approaches. This research addresses the challenge of preparing a data set to evaluate deep learning approach’s performance for the machine translation activity for Singlish to English language translation and to evaluate Seq2Seq Neural Machine Translation (NMT) model. The proposed seq2seq model is purely based on the attention mechanism, as it has been used to improve NMT by selectively focusing on parts of the source sentence during translation. The proposed approach can achieve 24.13 BLEU score on Singlish-English by seeing ~0.15 M parallel sentence pairs with ~50 K word vocabulary.