Learning the relationships between drug, symptom, and medical condition mentions in social media

Abstract

We consider the general problem of learning relationships between drugs, symptoms, and medical conditions mentioned on Twitter, with the goal of estimating probability distributions to reduce the difficulties presented by social media's incomplete picture. If a user mentions taking a drug and experiencing several unexpected symptoms, for example, are the symptoms associated with that drug or is it more likely that the symptoms are associated with an unmentioned underlying condition? We describe a model for learning from and utilizing such relationships. We demonstrate that our approach identifies drugs that are similar based on their associated symptoms (or conditions), identifies conditions that are similar based on their associated symptoms, and can determine whether a symptom is caused by a medical condition or by a drug (i.e., a drug side effect).