NLP and Deep Learning based Analysis of Social Media for Evaluating Public Sentiments

Societal perceptions regarding free speech regulations need advanced analytical methodologies to accurately assess the growing influence of social media within public discourse. SpeechRegNet uses Natural Language Processing (NLP) and Deep Learning to evaluate public sentiment through a sentiment-driven legal framework for analyzing free speech laws together with their interpretations and standpoints. The framework implements VADER (96.23%), BERT (98.82%), RoBERTa (97.94%), together with CNN-LSTM (95.41%) to achieve sentiment analysis and classification tasks. The stance detection system distinguishes between Pro-Free Speech and Against-Free Speech and Legal Perspective with successful outcomes reaching 98.21% for Pro-Free Speech and 97.36% for Public Opinion. User emotions receive Transformer-based classification that successfully identifies three emotions at rates of 97.56% (Joy), 98.73% (Fear) and 96.75% (Disgust). Additionally, legal text similarity analysis employs BERT embeddings (98.79%) and Legal-BERT (98.24%) to compare public sentiment with legislative texts. The F1-score result from SpeechRegNet evaluation shows 98.22% overall performance which validates its reliability in both sentiment and legal discourse evaluation. The research findings demonstrate meaningful variations between what people think and what legislators intend to do thus indicating the need for policy corrective measures. The results of this research enable policymakers to conduct regulatory impact analysis while uncovering misinformation which sets SpeechRegNet as a critical instrument for legal professionals as well as policymakers.