Deep Hybrid Bi-LSTM-CNN framework for medical condition and sentiment analysis for health prediction​

Biomedical and Therapeutics Letters

Read More

Keywords: item categories 

Section: Biomedical Technology
Keywords: Medical Sentiment Analysis, LSTM, Bi-LSTM, Machine Learning, Deep Learning, Medical AI

Medical and health related sentiment analysis has become a critical task in natural language processing (NLP) with applications in public health decision-making, patient opinion mining, and disease monitoring. However, challenges remain in accurately capturing context, polarity, and long-term dependencies in short and noisy texts such as tweets. This work designed a Deep Hybrid Bi-LSTM-CNN framework that integrates ELMo embeddings, CNN-based attention, and stacked Bi-LSTM layers for sentiment polarity detection in medical field. The model effectively captures both contextual and sequential dependencies while emphasizing sentiment-relevant features, enabling a nuanced interpretation of healthcare-related discourse. Experiments on a medical Twitter dataset demonstrate that the proposed framework achieves 97.2% accuracy, 94.5% sensitivity, and 98.1% specificity, outperforming baseline models including LSTM, Bi-LSTM, and CNN-LSTM by up to 6.8%. The results establish the robustness of the hybrid architecture for medical sentiment classification, underscoring its potential for real-world applications in healthcare monitoring, patient feedback analysis, and public health informatics.