A Deep Learning Approach to Extracting Adverse Drug Reactions

The wide use of Social media networks have changed the way patients share their health experiences. They offer valuable information on drugs and their side effects directly from patients. However, extracting useful information from social media sources is very challenging, due to several factors including grammatical and spelling errors, colloquial language, and post length limitation. This paper proposes a deep learning approach for extracting adverse drug reactions from twitter posts. It represents words as a vector of both domain and semantic features utilizing the rich medical terminology. The proposed method is evaluated on adverse drug events (ADRs) in tweets. Results show that the developed approach improves the precision of ADR detection by 15.28% over other state-of-the-art deep learning methods with a comparable recall score on twitter posts.