UCSYNLP-Lab Machine Translation Systems for WAT 2018.
Yi Mon Shwe Sin, Thazin Myint Oo, Hsu Myat Mo, Win Pa Pa, Khin Mar Soe, Ye Kyaw Thu · Institutional Repositories DataBase (IRDB) · 2018
In this description, we report the experimental results of Machine Translation models conducted by a team from University of Computer Studies, Yangon (UCSY) for the translation tasks of WAT 2018.Generally, our models are based on neural methods and statistical methods for both Myanmar-English and English-Myanmar direction of languages pair.For the neural method experiments, attention-based neural machine translation (NMT) that uses word level segmentation and Transformer that uses sub-word level segmentation have been carried out.In the portion of statistical machine translation (SMT), we used three different statistical approaches: phrase-based, hierarchical phrasebased, and the operation sequence model (OSM).Different Machine Translations are conducted on the ALT and UCSY datasets and the best scores from the experiments are described in this system description.