Asian Research Journal of Current Science

The rapid growth of social media platforms, online discussion forums and digital communication channels has generated large volumes of opinion-rich textual data that can support real-time public opinion monitoring. However, existing sentiment analysis systems often face limitations related to contextual ambiguity, scalability, computational overhead and noisy social media expressions. This study proposes a Cluster-Aware Deep Sentiment Mining (CADSM) framework for real-time public opinion monitoring. The framework integrates adaptive semantic clustering, distributed stream preprocessing, cluster-aware contextual feature augmentation and hybrid CNN-LSTM sentiment inference within a unified architecture. Opinion streams collected from X (formerly Twitter), Reddit, online discussion forums and news comment platforms were semantically grouped before distributed preprocessing and deep sentiment classification. Structural cluster metadata, including semantic density, cluster similarity, contextual relationships and sentiment distribution, were combined with textual embeddings to improve sentiment representation. The framework was evaluated using multilingual social media datasets containing positive, negative and neutral sentiment classes, and its performance was compared with SVM, Random Forest, CNN and CNN-LSTM models. Experimental results showed that CADSM achieved 97.4% accuracy, 96.9% precision, 96.5% recall and a 96.7% F1-score, with a false positive rate of 2.2%. The framework also achieved an average processing latency of 1.4 seconds and maintained accuracy above 96% as the stream volume increased from 10,000 to 500,000 messages. Under noisy conditions, CADSM maintained 92.3% accuracy at 5 dB SNR and achieved an AUC score of 0.98. These findings indicate that combining semantic clustering with distributed deep sentiment inference can improve contextual consistency, scalability and robustness for real-time public opinion monitoring while supporting timely sentiment interpretation across dynamic online discussions.