Window Classifiers and Conditional Random Fields for Medical Report De-Identification

JUAN MANUEL PÉREZ; FRANCO LUQUE; VIVIANA COTIK

Bilbao

Workshop; IberLEF 2019 Iberian Languages Evaluation Forum 2019; 2019

Iberian Languages Evaluation Forum

Information extraction of medical reports is key in order toimprove timely discoveries of findings and as an aid to improve decisionsabout medical treatments and budget. In order to develop informationextraction methods, medical data has to be available. Since this data isextremely sensitive due to the presence of personal information, reportde-identification is needed. We present two methods, a window classifierand an implementation of conditional random fields (CRF) in order tode-identify personal information of Spanish medical records provided bythe MEDDOCAN challenge. CRF obtained the best results with a F1-measure of 0.897 for named entity recognition with exact match (subtask1), and 0.930 and 0.940 for inexact match (subtask 2 strict and mergedrespectively).