Application of Neural Machine Translation with Attention Mechanism for Translation of Indonesian to Seram Language (Geser)

Abdul Wahid Rukua (Mathematic Department, Faculty of Mathematics and Natural Sciences, Universitas Pattimura, Indonesia)
Yopi Andry Lesnussa (Mathematic Department, Faculty of Mathematics and Natural Sciences, Universitas Pattimura, Indonesia)
Dorteus Lodewyik Rahakbauw (Mathematic Department, Faculty of Mathematics and Natural Sciences, Universitas Pattimura, Indonesia)
Berni Pebo Tomasouw (Mathematic Department, Faculty of Mathematics and Natural Sciences, Universitas Pattimura, Indonesia)

Pattimura International Journal of Mathematics (PIJMath)
Vol 2 No 2 (2023): Pattimura International Journal of Mathematics (PIJMath)

The Seram language (Geser) is one of the regional languages in Kabupaten Seram Bagian Timur of Maluku Province which has been classified by the Language Office as an endangered language. This study uses the Neural Machine Translation (NMT) method in an effort to preserve the Seram (Geser) language. The NMT method has proven to be effective compared to SMT in overcoming the challenges of language translation by using the attention mechanism to improve translation accuracy. The data used in this study were obtained through interviews of 3538 parallel corpus, 255 Indonesian vocabularies and 269 Seram (Geser) vocabularies. The result showed that using 708 test data without Out-of Vocabulary (OOV) the BLUE Score was 0.90518992895191 or 90.518%.