부산대학교 도서관

Textual sentiment analysis (TSA) has gained significant attention recently for its wide-ranging applications across various research domains and industries. However, most existing research and sentiment analysis tools are primarily tailored for English texts. The unique linguistic complexities of the Bengali language, coupled with a paucity of comprehensive resources and tools, pose distinctive challenges for TSA in Bengali. This paper introduces an intelligent approach, leveraging transformer-based learning techniques by harnessing the potent capabilities of self-attention mechanisms for dealing with Bengali sentences containing ungrammatical structures or local dialects. To tackle the downstream TSA task in Bengali, this work explores a range of machine learning (ML), deep learning (DL), and transformer-based baselines. Experimental results reveal that the Bangla BERT model outperforms the other baselines, achieving the highest weighted f 1 -score of 0.69.