TMH: A Transformer-Markov Hybrid Model for Behavior-Aware Code Summarization

Social media has emerged as a key platform in which people convey their opinions, emotions, and daily lives, with most of them showing cryptic indicators of mental disorders like depression. Detection of depression based on nonverbal behaviour has attracted significant attention. It is difficult to implement prior work on detecting depression in real-life situations since it was mainly focused on detecting depressed individuals in controlled laboratory environments. Depression reaches millions globally, and thus early and precise detection for timely intervention is essential. Diagnostic methods tend to be subjective and time-consuming based on the conventional methods. This paper proposes a multi-modal deep learning approach with hybrid remora optimization techniques (MDA-HRO) to integrate multi-modal data (video, audio, and text) to increase performance in emotional state of depression and sentiments. By combining the strengths of feature extraction techniques and a deep convolutional neural network optimized with hybrid optimization, the proposed framework effectively predicts sentiment across the modalities. The model is evaluated on the D-Vlog and D-Reddit datasets, achieving impressive accuracy rates for text data with 98%, audio data with 97%, and video data with 96%. Compared to conventional methods, the proposed work is more efficient and supports its applicability to various multimodal sentiment analysis tasks.