An Enhanced Multimodal Approach for Sentiment Analysis Using Deep Learning Techniques

Abstract

Social media has become a prominent medium through which individuals express their perspectives, sentiments, and experiences, frequently exhibiting apparent indications of mental health conditions such as depression. The identification of depression through nonverbal behaviour has garnered considerable interest. Implementing previous research concerning the detection of depression within real-world scenarios presents challenges, primarily because the research predominantly concentrated on identifying depressive individuals within controlled laboratory settings. Depression affects millions worldwide therefore, early and accurate detection is critical for prompt intervention. Conventional diagnostic methodologies are often subjective in nature and require considerable time to implement. This study introduces a hybrid remora-optimized multimodal deep learning
(HRO-MDL) approach to incorporate multi-modal data such as audio, video, and text for enhanced performance in the assessment of depressive emotional states and sentiments. The proposed framework accurately predicts sentiment across modalities by integrating feature extraction from various modalities using a deep convolutional neural network optimized with hybrid optimization. The model's performance was assessed using the D-Vlog and Depression Cleaned Reddit datasets, demonstrating notable accuracy levels 94.00% for text data, 92.00% for audio data, and 93.00% for video data. In contrast to traditional methodologies, the presented work exhibits enhanced efficiency and underscores its suitability for diverse multimodal sentiment analysis applications.