An empirical analysis on statistical and neural machine translation system for English to Mizo language

Machine Translation Systems for under-resource languages encounter quality and comprehension issues. Our research work focuses on the Statistical and Neural approaches methodologies for translating English into Mizo in a specific domain. We created an English-to-Mizo parallel dataset from the National Platform of Language Technology (NPLT) domains, the Bible and other domains as part of the system development. The performance of translations produced by Phrase-Based Statistical Machine Translation (PB-SMT) and Neural Machine Translation (NMT) systems were trained and tested in under-resource and domain-specific circumstances which were then explored thoroughly utilizing automatic and subjective evaluation approaches. The experiment conducted with PB-SMT displayed better results as compared to the state-of-the-art NMT on English to Mizo translation works. The testing quality of our system was evaluated through a suitable example with automatic BLEU and Manual evaluation consisting of two parameters namely, adequacy and fluency. © 2023, The Author(s), under exclusive licence to Bharati Vidyapeeth's Institute of Computer Applications and Management.