<![CDATA[The development of Artificial Intelligence (AI) has generated diverse public responses, which are widely expressed through social media platforms such as Twitter/X, TikTok, and YouTube. Understanding public sentiment toward AI is important because it provides insights into public acceptance, concerns, and perceptions regarding the development of this technology. However, sentiment analysis of Indonesian-language social media text remains challenging due to the informal and unstructured nature of the language, including slang, typographical errors, and substantial morphological variations. This study aims to develop a hybrid sentiment analysis model based on Bidirectional Long Short-Term Memory–Conditional Random Fields (BiLSTM–CRF), integrating Word2Vec and FastText to improve the performance of public sentiment classification toward Artificial Intelligence. The dataset was collected from three social media platforms—Twitter/X, TikTok, and YouTube—and comprised 30,550 comments after preprocessing and labeling into three sentiment categories: positive, neutral, and negative. The experiments compared the proposed models with the baseline CRF, LSTM, and BiLSTM models using accuracy, precision, recall, and F1-score as evaluation metrics. The results show that the BiLSTM–CRF model with FastText achieved the best performance, with an accuracy of 93.12%, precision of 92.74%, recall of 92.31%, and F1-score of 92.52%. This performance exceeded that of BiLSTM–CRF with Word2Vec (91.38%), BiLSTM (87.24%), LSTM (84.15%), and CRF (78.42%). These findings demonstrate that integrating BiLSTM–CRF with FastText's subword-based representation is effective in improving sentiment analysis performance on Indonesian-language social media data. This study contributes to the development of more robust, adaptive, and representative sentiment analysis methods for understanding public opinion toward Artificial Intelligence.]]>
Copyrights © 2026
