Skip to main content
. Author manuscript; available in PMC: 2017 Apr 1.
Published in final edited form as: J Biomed Inform. 2016 Jan 13;60:14–22. doi: 10.1016/j.jbi.2016.01.003

Table 4.

Features used to develop the classification model.

Feature
Category		 Feature Type Number of
features		 Description
Semantic Predication 7 Total number of predications with a treatment-related predicate (1 feature) and number of predication instances per treatment-related predicate (i.e., TREATS / NEG_TREATS, ADMINISTERED_TO / NEG_ADMINISTERED_TO, AFFECTS / NEG_AFFECTS, PROCESS_OF / NEG_PROCESS_OF, PREVENTS / NEG_PREVENTS, and COMPARED_WITH / HIGHER_THAN / LOWER_THAN / SAME_AS) - 6 features.
Semantic Population 1 Whether or not a sentence includes a description of the types of patients who are eligible to receive a certain treatment.
Semantic Concept 5 Total number of concepts in the sentence (1 feature) and number of concept instances per UMLS semantic group (i.e., Chemicals & Drugs (CHEM), procedures (PROC), physiology (PHYS), and disorders (DISO) - 4 features.
Syntactic Text-based 4 Four categories of potentially useful cue terms mentioned in Table 3 including “References to external content”, “Study design”, “Deontic terms and evidence source attributions” and “Treatment”.