Long Short Term Memory (LSTM) Improvement Accuracy Using FastText and Glove for Language Translation Batak - Indonesian

This research discusses the improvement of Long Short-Term Memory (LSTM) model accuracy in translating Batak to Indonesian by utilizing FastText and GloVe embeddings. Traditional LSTM models often struggle with the complexity and richness of underrepresented languages like Batak. By integrating FastText and GloVe, which provide word vectors with deeper semantic meanings, the translation quality can be significantly enhanced. Experimental results show that this approach offers a substantial performance improvement compared to conventional LSTM models, producing more accurate and contextual translations. These findings have the potential to be applied in broader multilingual translation systems and support the preservation and digital accessibility of minority languages.