Skip to main content

Table 3 Comparison of event extraction performance

From: Unsupervised inference of implicit biomedical events using context triggers

Existing models  F1RecallPrecision
BB-event task participants (2016)LIMSI48.564.638.8
 TurkuNLP52.144.862.3
 VERSE55.861.551.0
State-of-the-art systemsLi et al. [23]58.158.056.3
 Li et al. [24]57.456.859.4
 Gupta et al. [20]58.765.753.0
Proposed models#intra#crossF1RecallPrecision
M1: Intra-clause syntactic patterns213037.730.748.8
M2: Intra-clause syntactic patterns + trigger-based inference (train)246040.434.848.1
M3: Intra-clause syntactic patterns + trigger-based inference (unlabeled)4176456.768.648.2
M4: VERSE (2016) + trigger-based inference (unlabeled)3395458.963.854.9
M5: VERSE (2016) + trigger-based inference (unlabeled) (without linguistic modality detection)3396358.363.853.6
  1. The numbers in boldface indicate the highest scores for each metric