Classifying Drug Ratings Using User Reviews with Transformer-Based Language Models

Abstract

Social web contains a large amount of information with user sentiment and opinions across different fields. For example, drugs.com provides users' textual review and numeric ratings of drugs. However, text reviews may not always be consistent with the numeric ratings. In this project, we built different classification models to classify user ratings of drugs with their textual review. Multiple supervised machine learning models including Random Forest and Naive Bayesian classifiers were built with drug reviews using TF-IDF features as input. Also, transformer-based neural network models including BERT, BioBERT, RoBERTa, XLNet, ELECTRA, and ALBERT were built for classification using the raw text as input. Overall, BioBERT model outperformed the other models with the overall accuracy of 87%. This research demonstrated that transformer-based classification models can be used to classify drug reviews and identify reviews that are inconsistent with the ratings.