Effect of mono corpus quantity on statistical machine translation Indonesian-Lampung dialect of nyo

Abstract

Lampung Province is located on the island of Sumatera. For the immigrants in Lampung, they have difficulty in communicating with the indigenous people of Lampung. As an alternative, both immigrants and the indigenous people of Lampung speak Indonesian. This research aims to build a language model from Indonesian language and a translation model from the Lampung language dialect of nyo, both models will be combined in a Moses decoder. This research focuses on observing the effect of adding mono corpus to the experimental statistical machine translation of Indonesian-Lampung dialect of nyo. This research uses 3000 pair parallel corpus in Indonesia language and Lampung language dialect of nyo as source language and uses 3000 mono corpus sentences in Lampung language dialect of nyo as target language. The results showed that the accuracy value in bilingual evalution under-study score when using 1000 sentences, 2000 sentences, 3000 sentences mono corpus show the accuracy value of the bilingual evaluation under-study, respectively, namely 40.97 %, 41.80 % and 45.26 %.