Sentiment Analysis of Social Media Using Deep Learning and Explainable AI

Sentiment analysis has become a fundamental task in Natural Language Processing (NLP), especially with the exponential growth of social media data. This paper presents a comprehensive approach for sentiment classification using multiple models including VADER, LSTM, BERT, and RoBERTa. The proposed system focuses on fine-tuning RoBERTa to achieve high accuracy and integrates Explainable Artificial Intelligence (XAI) using DeepSeek to enhance interpretability. Experimental results show that the proposed model achieves an accuracy of approximately 95–96%, outperforming baseline models. The integration of DeepSeek provides human-readable explanations, improving model transparency and usability.