Enhancing Aspect-Based Sentiment Analysis for Indian Hospital Reviews: A Weighted Average Ensemble Approach

Public reviews are essential for judging hospital effectiveness and ensuring patients receive the best medical care. Aspect-based sentiment analysis handles sentiment analysis by splitting it into parts. It struggles to find sentiments when inputs are made up of languages, and there are a lot of noisy elements. It is designed to improve how well aspects are classified using ensemble and machine learning. We used conditional random fields (CRF) and the dataset of 29,752 hospital reviews to identify the elements. After that, we worked with Support Vector Machine, K-Nearest Neighbours, and Naïve Bayes using them for classification tasks. Weighted Average Ensemble and AT features were combined, and then a 5-fold cross-validation with a micro F1-score followed. Tuning hyperparameters through optimisation made each classifier function better. Once the improved ensemble was used, the system scored 0.8441 on the F1-score, better than all other models and classifiers in the experiment. This research indicates that using a combination of approaches is helpful for domain-specific sentiment analysis, and these techniques could significantly improve health field feedback interpretation and drive further progress in the field.