OptiASAR: Optimized Aspect-Based Sentiment Analysis of Reviews With BiLSTM-GRU and NER-BERT in Healthcare Decision-Making

In the healthcare sector, understanding patient feedback through Aspect-Based Sentiment Analysis (ABSA) is vital, as it provides detailed insights into various elements like doctor-patient interactions, treatment efficacy, and hospital amenities. Traditional sentiment analysis methods may not capture the diverse aspects and often miss these subtleties. This paper proposes an OptiASAR (Optimized Aspect-based Sentiment Analysis of Reviews) model integrating a Bidirectional Long Short-Term Memory-Gated Recurrent Unit (BiLSTM-GRU) for enhanced sequence modeling, improving the capture of long-term dependencies in healthcare reviews. By using the Named Entity Recognition and Bidirectional Encoder Representations from Transformers (NER-BERT) embeddings, the model enhances the representation and contextual comprehension of aspect terms. Furthermore, a multi-headed self-attention mechanism is employed to focus on different parts of the input, thereby capturing complex relationships and dependencies more effectively. The study highlights significant improvements in sentiment analysis performance, particularly within the healthcare domain. The proposed OptiASAR model achieved promising performance in terms of accuracy compared with contemporary models, offering healthcare professionals and stakeholders a robust tool to understand sentiments related to various healthcare service aspects, thus, aiding informed decision-making and quality enhancement initiatives.