Towards Automatic Pharmacovigilance: Analysing Patient Reviews and Sentiment on Oncological Drugs

Abstract

The collection, detection and monitoring of information such as side effects, adverse effects, warnings, precautions of pharmaceutical products is a challenging task. With the advent of user forums, online reviews have become a significant source of information about products. In this work, we aim to utilize pharmaceutical drugs reviews by patients on various health communities to identify frequently occurring issues. We compare these issues with food and drug administration (FDA) approved drug labels for possible improvements. We focus on Oncological drugs and develop a scalable system for mapping of interventions against indication and the respective symptoms from patient comments. Using these mappings, our system is able to compare different sections of FDA labels for recommendations. We use SVM based framework for sentiment analysis to give an overall rating to the drugs. We further incorporate aspect based sentiment analysis for finding the orientation of drug reviews for specific targets.