Knowledge-based Feature Engineering for Detecting Medication and Adverse Drug Events from Electronic Health Records
Proceedings of the 1st International Workshop on Medication and Adverse Drug Event Detection, PMLR 90:31-38, 2018.
This paper presents a Conditional Random Field (CRF) learning model integrated with a series of knowledge-based feature engineering functions for detecting medication and adverse drug events from electronic health records (EHRs). Our experimental evaluation shows high performance in terms of the F-score measure (83.4% and 92.5% respectively for strict and relax modes) in the detection of medication and clinical finding named entities. It also shows promising results (overall 78.8% F-score) in the recognition of detailed properties of medication use as well as different types of clinical findings mentioned in electronic health records.