Skip to Main Content
Table 5 

There are two classes of feature templates: lexical features (TriggerPred,Trace*) and non-lexical features (PredRel,PredRelPred). The lexical features are relatively much more important for obtaining good accuracy (76.4% vs. 23.1%), but adding the non-lexical features makes a significant contribution as well (84.7% vs. 76.4%).

Features
Accuracy (%)
PRED 13.4 ±1.6 
PRED + PREDREL 18.4 ±3.5 
PRED + PREDREL + PREDRELPRED 23.1 ±5.0 
PRED + TRIGGERPRED 61.3 ±1.1 
PRED + TRIGGERPRED + TRACE* 76.4 ±2.3 
PRED + PREDREL + PREDRELPRED + TRIGGERPRED + TRACE* 84.7 ±3.5 
Features
Accuracy (%)
PRED 13.4 ±1.6 
PRED + PREDREL 18.4 ±3.5 
PRED + PREDREL + PREDRELPRED 23.1 ±5.0 
PRED + TRIGGERPRED 61.3 ±1.1 
PRED + TRIGGERPRED + TRACE* 76.4 ±2.3 
PRED + PREDREL + PREDRELPRED + TRIGGERPRED + TRACE* 84.7 ±3.5 
Close Modal

or Create an Account

Close Modal
Close Modal