Hybrid system for adverse drug event detection

[edit]

Alec B. Chapman, Kelly S. Peterson, Patrick R. Alba, Scott L. DuVall, Olga V. Patterson ;
Proceedings of the 1st International Workshop on Medication and Adverse Drug Event Detection, PMLR 90:16-24, 2018.

Abstract

In context of the NLP Challenges for Detecting Medication and Adverse Drug Events from Electronic Health Records (MADE 1.0) \citep{Yu}, we built a hybrid natural language processing system that combined multiple algorithms and resources to identify the relationship between mentions of symptoms and drugs. Our system employed a conditional random field (CRF) model for named entity recognition (NER) and a random forest model for relation extraction (RE). Final performance of each model was evaluated separately and then combined on the challenge’s hold-out evaluation set. The micro-averaged F1 score was 80.9% for NER, 86.8% for RE, and 59.2% for the final system.

Related Material