Evaluating Consumer Feedback on Drug Effects: A Comparative Study of Sentiment Analysis Models and Visual Insights

Abstract

Introduction:

Aspect-based sentiment analysis (ABSA) has emerged as a significant technique for comprehending patient feedback on specific attributes of medications, including effectiveness, side effects, ease of use, cost, and customer service [1]. This study investigates the application of deep learning models, particularly BERT, for ABSA of drug reviews. BERT is fine-tuned for each aspect and its performance is benchmarked against other state-of-the-art models such as RoBERTa, LSTM, and traditional machine learning approaches [2]. The findings show that BERT outperforms SVM, Naive Bayes, and LSTM models in a number of parameters, including accuracy, precision, recall, and F1-score [3]. This study highlights ABSA's capacity to offer deep insights into patient experiences, which can be applied to improve medication quality and patient satisfaction[4].

This study also includes data visualization methods to improve the understanding of ABSA results. An interactive platform has been created where users can choose a specific medication and see the sentiment distribution across different aspects. This enables a thorough analysis of patient opinions and trends. By combining ABSA with data visualization, we can gain actionable insights that assist healthcare professionals, pharmaceutical companies, and policymakers in making informed decisions. Future research could investigate more advanced visualization techniques and model enhancements for wider applications.