2026-04-16T03:28:59Z https://meral.edu.mm/oai

oai:meral.edu.mm:recid/00007669 2021-12-13T04:13:14Z 1582963674932 1582963674932:1597396989070 user-miit

Neural Machine Translation between Myanmar Sign Language and Myanmar Written Text Swe Zin Moe Ye Kyaw Thu Hnin Aye Thant Nandar Win Min We explore Neural Machine Translation (NMT) between Myanmar Sign Language (MSL) and Myanmar Written Text (MWT). Our developing MSL-MWT parallel corpus was used and the experiments were carried out using three different NMT approaches: Recurrent Neural Network (RNN), Trasformer, and the Convolutional Neural Network (CNN). In addition, four different segmentation schemes for word embedding were studies, these were syllable segmentation, word segmentation (sign unit based word segmentation for MSL), SentencePiece and the Byte-Pair-Encoding (BPE). The results show that the highest quality NMT and Statistical Machine Translation (SMT) performances were attained with syllable segmentation for both MSL and MWT. We found that Transformer outperformed both CNN and RNN for MWT-to-MSL and MSLto-MWT translation tasks. 2018-12-18 http://hdl.handle.net/20.500.12678/0000007669 https://meral.edu.mm/records/7669