Deep Learning Approach for Aspect-Based Sentiment Analysis on Indonesian Hospitals Reviews

Hospitals are individual health service facilities that provide inpatient and outpatient care, therefore, quality hospital services, facilities, and resources are a necessity and must be met by every hospital. To assess the quality of hospital services, facilities, and resources, we can see reviews from users through reviews available on Google Maps. Using natural language understanding, we can identify the sentiment associated with each particular aspect. In this research, we developed a deep learning model which can classify three aspects (services, human resources, and facilities) with four sentiments (positive, negative, neutral, and none) using a multiclass-multioutput deep learning model, ensuring both aspect and sentiment classification within a single model. Our research concludes that the BERT-LSTM model demonstrates promising performance with an F1-score of 0.597, but faces challenges in accurately classifying neutral sentiment across all aspects due to the lack of neutral sentiment labels.