2026-06-25T02:02:50Z https://meral.edu.mm/oai

oai:meral.edu.mm:recid/4685 2021-12-13T04:14:49Z 1582963302567:1597824273898 user-ucsy

Syllable-based Myanmar-English Neural Machine Translation Sin, Yi Mon Shwe Soe, Khin Mar The paper presents the first large scaleevaluation of the quality of Syllable-based NeuralMachine Translation (Syllable-NMT) system forMyanmar-English pair. Neural Machine Translation(NMT) system has reached state-of-the-arts resultson some languages. However, one of the mainchallenges that NMT still faces is dealing with verylarge vocabularies and morphologically richlanguages. Like other low-resources languages,Myanmar Language has a lots of morphologyinformation. This issue lead is to increase theambiguity and to decrease the quality of translationresults. Moreover, rule-based and phrase-basedtechniques were used in the existing research onMyanmar translation with the small amount ofparallel corpus. Therefore, a large amount ofparallel corpus is prepared and introduces a NMTmodel that maps a source syllable sequence to atarget word sequences to address the morphologicalproblems. In addition, this paper shows someexperiments results and compare them. Our resultsshow that syllable-NMT system is able to surpassthan the character-based and word-based NMTsystems by 5 BLEU. 2018-02-22 http://hdl.handle.net/20.500.12678/0000004685 https://meral.edu.mm/records/4685