Detecting Adverse Drug Reactions from Social Media Based on Multichannel Convolutional Neural Networks Modified by Support Vector Machine

Abstract

Prescribing medication is a task that physicians face every day. However, when prescribing medication, physicians should be aware of all possible side effects of the drug. Drug-related side effects or Adverse Drug Reactions (ADR) may have profound effects on patients' quality of life as well as putting more pressure on the health care system. Due to the complexity of the diagnosis process, there are still a number of important unknown ADRs. Social media collects large amounts of information about drug use from patients, therefore may be a useful way for extracting ADRs. As a result, the social media becomes one of the effective tool for ADR because users share their experiences and opinions in different fields every day, such as their health, unknown side effects of a drug, and so on. In this study, we propose a new method for identifying ADRs. To meet the challenge of displaying data from multiple sources as well as identifying text containing drug reaction information, a new deep learning architecture is proposed which is based on multichannel convolutional neural networks. The obtained results from applying the proposed architecture on real data obtained from Twitter demonstrates its potential in recognizing ADRs.