<![CDATA[Bengkulu Malay is a regional language in Indonesia widely used in digital communication, particularly on social media. Comments in Bengkulu Malay reflect various sentiments toward local issues; however, sentiment analysis remains challenging due to limited data and unique linguistic characteristics. This study aims to evaluate the performance of three deep learning architectures—Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Bidirectional LSTM (Bi-LSTM)—in classifying the sentiment of Bengkulu Malay text comments. The three models were tested using specific preprocessing approaches, including tokenization and stopword removal tailored to the structure of the Bengkulu Malay language. The results indicate that the Bi-LSTM model outperforms both CNN and LSTM in analyzing the sentiment of Bengkulu Malay comments. This is evidenced by the consistent dominance of Bi-LSTM across all datasets, achieving a peak accuracy of 91% on the Q1 dataset, while the LSTM and CNN models lagged behind by a significant margin. Given its stable performance and superior accuracy, the Bi-LSTM model proves to be the most effective architecture and is recommended for sentiment analysis of Bengkulu Malay comments. ]]>
Copyrights © 2026
