Machine Learning-Based Komering Language Translation Engine with Bidirectional RNN Model Algorithm

Regional languages represent a significant facet of Indonesian culture, spanning various regions across the archipelago. However, the knowledge of these regional languages is often limited among individuals outside their respective regions. For instance, the Komering Language remains relatively unfamiliar to the inhabitants of the South Sumatra region. The cultural significance of regional languages is at risk of diminishing due to unfamiliarity with the associated words and sentences. Therefore, there is a pressing need to preserve local language cultures to prevent their marginalization. In this research, the authors aim to develop a bidirectional Recurrent Neural Network (RNN) model capable of translating Indonesian to the Komering language using machine learning techniques. In the study, a bidirectional model was evaluated using various types of RNN, namely simpleRNN, GRU, and LSTM. The results showed that the bidirectional model with simpleRNN yielded a performance accuracy of 87.12% after 100 epochs.