Automated Clinical Text Categorization and Sentimental Analysis

Healthcare institutions generate vast amounts of unstructured clinical text, making automated analysis essential for efficient decision-making and improved healthcare outcomes. This study proposes a transformer-based approach utilizing Bidirectional Encoder Representations from Transformers (BERT) for both sentiment analysis and medical specialty classification of electronic health records, including clinical notes and discharge summaries. To address class imbalance and improve model performance, the study focuses on the most dominant specialty categories within the dataset. The proposed framework leverages BERT's ability to capture deep contextual and semantic relationships in medical narratives, enabling accurate classification of clinical content as well as effective detection of sentiment polarity. The model is evaluated using standard performance metrics and demonstrates superior results compared to traditional machine learning approaches. The findings highlight the effectiveness of transformer-based models in handling complex medical text and provide a scalable solution for automated clinical text analysis. This approach facilitates faster information extraction, reduces manual workload, and supports enhanced clinical decisionmaking and patient care.