An Enhanced Hybrid Framework for Scalable Multi-Modal Sentiment Analysis with Text Summarization

In today's digital age, sentiment analysis from vast and diverse data streams presents a significant challenge. Traditional methods focus primarily on textual inputs, often overlooking valuable information embedded in visual content. This paper proposes a novel hybrid model that integrates text summarization with multi-modal sentiment analysis, leveraging BERT and RoBERTa for textual data and ResNet for image data. By summarizing and fusing information from both modalities, the model captures nuanced sentiment representations. Evaluations on large-scale datasets, including Twitter-1517, demonstrate improved performance over unimodal and baseline multimodal models. This approach paves the way for scalable, efficient sentiment analysis across platforms rich in both text and visuals.