Sentiment Analysis of Sinhala Patient Feedback using Deep Learning Techniques

Abstract

Patient feedback has become an essential tool for evaluating and improving healthcare service quality. However, low-resource languages, such as Sinhala, have been largely overlooked in sentiment anal ysis research, which has mostly focused on widely used languages like English. This study addresses this gap by developing a deep learning-based sentiment analysis system for Sinhala patient feedback in hospital settings. A dataset of Sinhala patient reviews was collected and manually annotated into three sentiment categories: positive, negative, and neutral. The text data underwent preprocessing steps, including cleaning, tokenization, and padding, to prepare it for deep learning models. Multiple architectures were implemented and evaluated, including Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory networks (BiLSTM). Performance was assessed using accuracy, precision, recall, and F1-score. The BiLSTM model achieved the highest accuracy of 98.80%, outperforming the CNN model, which recorded 97.60% ac curacy. These results demonstrate that deep learning models can effectively capture sentiment patterns in Sinhala text, even with limited resources. This study highlights the potential of deep learning techniques to provide actionable insights into patient satisfaction, enabling hospitals to identify areas for service improve ment. In addition, the work contributes to advancing natural language processing research for low-resource languages by demonstrating that high-performing sentiment analysis models can be developed with careful annotation and preprocessing. The findings suggest that integrating such systems into healthcare workflows could support data-driven decision-making and improve patient experiences. Overall, this research not only offers a practical tool for analyzing Sinhala patient feedback but also serves as a foundation for further studies in low-resource language NLP applications.