Improving English-Vietnamese neural machine translation with linguistic factors

Abstract

Deep learning has been shown to be successful in a number of domains of natural language processing, especially neural machine translation (NMT). Though promising, NMT still lacks the ability of modeling deeper semantic and syntactic aspects of the language. In this paper, explore different linguistic annotations at the word level, including: part-of-speech, named entity, and word cluster to integrate into the NMT framework. Experiments show that adding these linguistic factors will help the NMT models in reducing language ambiguity or alleviating data sparseness problems.