Multimodal Affective Communication Analysis: Integrating Text Sentiment with Speech Emotion using Transfer Learning

The final aim is to integrate sentiment analysis and speech emotion recognition through transfer learning. These traditional systems analyse text and speech separately - thereby missing a lot of emotional expressions that cannot be captured in either text or speech alone. Hence, the presented work integrates both modalities to improve emotion detection accuracy. Advanced natural language processing models complete the task of text sentiment analysis while features from acoustics and deep learning techniques help complete the task of speech emotion recognition. Its power with transfer learning – leveraging. These methods will primarily improve the final recognition performance of speech emotion recognition where available labelled data can't measure up with their original corpus. The study indicates that such late fusion techniques compared with unimodal systems work way better- more robust and accurate class recognition of emotions is achievable.