Performance Analysis of GloVe Word Embedding for Batak – English Translation

This paper presents a performance analysis of GloVe word embeddings in Batak language – English, utilizing Long Short-Term Memory (LSTM) models. The Batak language, with its rich language features and relatively low resource availability, poses significant challenges for machine translation tasks. By integrating GloVe embeddings into LSTM networks, we aim to enhance the quality and accuracy of translation outputs. The study compares the effectiveness of GloVe embeddings against Word2Vec embeddings. Our experiments involve training and evaluating LSTM models with both embedding techniques, followed by an analysis of translation performance metrics. The result demonstrate the relative strengths and weakness of each embedding approach, providing valuable insight into their suitability for low-resource language translation tasks.