LSTM-Based Machine Translation for Madurese-Indonesian

Madurese is one of the regional languages in Indonesia, which dominates East Java and Madura Island particular. The use as a daily language has declined significantly due to shift children adolescents, some are caused by sense prestige difficulty learning Madurese. scarcity research or scientific titles that raises also helps reduce literacy language. Our focuses on creating translation machine for Indonesian maintain preserve existence so can be done through digital media. This study latest dataset Madurese-Indonesian using corpus 30,000 Madura-Indonesian sentence pairs from online Bible. scrapped Bible pages organize based bilingual Then manually process text match two languages' scrapping results, normalization, tokenization remove non-printable characters punctuation corpus. To perform neural (NMT), connected RNN encoder with decoder model, while training testing, used sequential model LSTM, BLEU measure was assess accuracy results. SoftMax optimization function Adam Optimizer added settings, including 128 layers adding Dropout layer got average evaluation result BLEU-1 0.798068, BLEU-2 0.680932, BLEU-3 0.623489, BLEU-4 0.523546 five tests conducted. Given differences between Indonesian, this best approach