Recognizing Mentions of Adverse Drug Reaction in Social Media Using Knowledge-Infused Recurrent Models

Gabriel Stanovsky,Daniel Gruhl,Pablo Mendes

doi:10.18653/v1/e17-1014

Abstract

Recognizing mentions of Adverse Drug Reactions (ADR) in social media is challenging: ADR mentions are context-dependent and include long, varied and unconventional descriptions as compared to more formal medical symptom terminology. We use the CADEC corpus to train a recurrent neural network (RNN) transducer, integrated with knowledge graph embeddings of DBpedia, and show the resulting model to be highly accurate (93.4 F1). Furthermore, even when lacking high quality expert annotations, we show that by employing an active learning technique and using purpose built annotation tools, we can train the RNN to perform well (83.9 F1).

Highlights

Identifying medical concepts in social media narratives is the task of recognizing certain phrases in the context of a user’s post
In this work we focus on the identification of Adverse Drug Reactions (ADR)
CSIRO Adverse Drug Event Corpus (CADEC) used brat (Stenetorp et al, 2012) to annotate five types of medical concepts: (1) Drug, names of medicine or drug, e.g., “Diclofenac” or “Aspirin”; (2) Adverse Drug Reaction, an unwanted reaction which according to the text is clearly associated with taking the drug, e.g, “acute stomach pain”; (3) Disease, the reason for taking the drug, e.g., “insomnia” or “aggression”; 1http://www.askapatient.com

Summary

Introduction

Identifying medical concepts in social media narratives is the task of recognizing certain phrases in the context of a user’s post. The recently created CSIRO Adverse Drug Event Corpus (CADEC) (Karimi et al, 2015) contains medical concepts annotation in posts from Ask a Patient, an online forum collecting medical patient narratives. CADEC used brat (Stenetorp et al, 2012) to annotate five types of medical concepts: (1) Drug, names of medicine or drug, e.g., “Diclofenac” or “Aspirin”; (2) Adverse Drug Reaction, an unwanted reaction which according to the text is clearly associated with taking the drug, e.g, “acute stomach pain”; (3) Disease, the reason for taking the drug, e.g., “insomnia” or “aggression”; 1http://www.askapatient.com (4) Symptom, manifestations of the disease, e.g., “trouble sleeping” or “constantly angry”; and (5) Finding, a clinical finding that does not pertain to any of the above categories.

Methods

Results

Conclusion