Advancing Medical Diagnosis: Enhancing Sentiment Analysis in Electronic Medical Records with Transformer Models

Abstract

This study investigated the application of the Transformer model for sentiment analysis on electronic medical records (EMRs) as a supportive tool for medical diagnosis. By conducting extensive comparative experiments with traditional deep learning models such as Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), and Bidirectional LSTM (BiLSTM), the findings revealed that the Transformer model consistently outperformed these models in sentiment classification tasks. Notably, the Transformer achieved higher accuracy and F1 scores, indicating its superior ability to capture contextual nuances within EMR data. The proposed method not only enhances the recognition accuracy of sentiment information embedded in medical records but also offers a novel approach to EMR analysis and sentiment detection. This advancement holds substantial promise for intelligent applications in healthcare, providing practitioners with valuable insights and contributing to more precise diagnostic support through advanced sentiment recognition capabilities in medical data processing.