Bengali Speech Sentiment Analysis Using Machine Learning Models: A Comparative Study
Mohammad Tanveer Shams, Md. Akib Hasan, Animesh Das Chowdhury, Tabassum Jahan Lamia, Md. Reasad Zaman Chowdhury, Mohammad Marufur Rahman · 2024
Though scholars find speech sentiment analysis based on audio data to be a very intriguing study topic, not enough work has been done for the fifth most spoken language in the world, Bangla. The purpose of this study is to close this research gap. SUBESCO, BanslaSER, and KBES combined dataset were utilized in this study's evaluation of all the models. Evaluations have been done on CNN models, ML models, and sequence models like LSTM and BI-LSTM. Mel-frequency spectrum was exploited by CNN models, while machine learning and sequence models were applied to numerical features. DenseNet201 achieved the finest accuracy of all the models, at 94%o. Finally, utilizing both hard and soft voting, all of the models based on numerical features and spectrogram features were ensembled, yielding accuracy rates of 94.99%o for numerical features and 95% for spectrogram features.