Multi-Modal Sentiment Analysis of Social Media Using CNN-LSTM Hybrid Models

Abstract

Social media platforms generate massive volumes of content in multiple formats, such as text, images, and videos, reflecting users’ opinions and emotions. Conventional sentiment analysis methods often focus solely on textual data, ignoring valuable cues present in visual content. This study presents a CNN-LSTM hybrid model for multi-modal sentiment analysis, leveraging Convolutional Neural Networks (CNNs) to extract features from images and Long Short-Term Memory (LSTM) networks to model sequential dependencies in text. The proposed approach fuses textual and visual features to enhance sentiment classification into positive, negative, and neutral categories. Evaluation on a multi-modal social media dataset demonstrates that the hybrid model significantly outperforms single-modality models in terms of accuracy, precision, recall, and F1-score. The findings underscore the importance of integrating multiple data modalities for robust sentiment prediction, with potential applications in brand monitoring, social media analytics, and public opinion tracking.