Low-Resource Single-Domain Machine Translation untuk Bahasa Karo-Indonesia

Abstract

Indonesia has cultural and linguistic diversity that must be preserved. One way is to utilize currently developing technology, Neural Machine Translation, to build a translation machine. This research uses a Statistical Machine Translation (SMT) approach. This computer-based translation approach is an approach based on statistical methods introduced since the mid-20th century. The popularity of SMT arises due to a translation architecture based on strong mathematical theory, good quality translation in many test runs, and the existence of toolkits that can be used to build MT models in a short time. In this research, the author built a translation machine from Indonesian to Karo Batak language using the Karo regional language Bible domain (Pustaka Sibadia) as the Karo language corpus and the Contemporary Indonesian Bible (BIMK) as the Indonesian language corpus. The best blue score obtained in this study was 0.18. More in-depth research is needed to maximize the results