Labeling events such as Adverse Drug Events in medical documents is critical for healthcare applications but challenging due to the variability in how events are documented in clinical notes. As part of a top performing submission to the NLP Challenges for Detecting Medication and Adverse Drug Events (MADE 1.0) we propose a model which leverages word embeddings and medical ontologies with performance comparable to Recurrent Neural Networks without requiring GPUs given its much faster training.
Learning Objective 1: Readers will learn alternative approaches to deep neural networks, which can be developed on more flexible hardware.
Kelly Peterson (Presenter)
VA Salt Lake City Health Care System
Alec Chapman, VA Salt Lake City Health Care System
Patrick Alba, VA Salt Lake City Health Care System
Scott DuVall, VA Salt Lake City Health Care System
Olga Patterson, VA Salt Lake City Health Care System