Tac srie 2018: Extracting systematic review information with medacy

Abstract

This draft paper describes our participation in the TAC SRIE 2018 Challenge for extracting experimental design factors for the systematic review process from the method section of biomedical journal articles. We discuss our system, medaCy, a python-based supervised multi-class classification system that uses Conditional Random Fields. Our system place fourth for the challenge. The results show that we are better able to identify Animal and Dose Group design factors than Exposure and Endpoint. This paper includes an analysis of our results and future directions to improve the extraction.

Publication
Text Analysis Conference (TAC) Systematic Review Information Extraction (SRIE)