Identification and Progression of Heart Disease Risk Factors in Diabetic Patients from Longitudinal Electronic Health Records

Table 2

Features used by smoking history, sectionizer, and time attribute assigner classifiers.

Component	Classification	Classifier	Classes	List of features
Smoking history	Sentence level	Naïve Bayes	Current, past, and never	Bag of words, POS tags
Sectionizer	Sentence level	Conditional random fields	Section heading, section heading with text, and text	First word uppercased, all words uppercased, all words lowercased, dictionary match, first word, second word, previous sentence features, next sentence features, full stop, and containing colon
Time attribute assigner	Phrase level	Naïve Bayes	Before DCT, during DCT, after DCT, and continuing	Identified risk factor spans, previous word, previous word POS tag, next word, next word POS tag, section information, and indicator attribute