Skip to main content

Advertisement

Table 2 Stacking experiments on Genia

From: Combining joint models for biomedical event extraction

System Simple Binding Regulation Overall
UMass 74.7 47.7 42.8 54.8
Stanford (1N) 71.3 44.2 35.6 49.9
Stanford (1P) 70.7 42.6 34.6 49.1
Stanford (2N) 69.0 40.5 30.8 46.5
Stanford (2P) 72.0 40.3 35.4 49.5
Stanford (1N, reranked) 71.7 46.9 35.4 50.2
Stanford (1P, reranked) 70.6 44.5 35.5 49.4
Stanford (2N, reranked) 70.2 44.6 38.2 47.9
Stanford (2P, reranked) 71.1 47.8 35.9 50.5
Stanford (all, reranked) 71.6 48.1 35.9 50.7
UMass←Stanford (all) (= FAUST) 76.9 43.5 44.0 55.9
UMass←Stanford (1N) 76.4 45.1 43.8 55.6
UMass←Stanford (1P) 75.8 43.1 44.6 55.7
UMass←Stanford (2N) 74.9 42.8 43.8 54.9
UMass←Stanford (2P) 75.7 46.0 44.1 55.7
UMass←Stanford (all, anchors) 76.4 41.2 43.1 54.9
UMass←Stanford (all, arguments) 76.1 41.7 43.6 55.1
UMass←Stanford (all) without novel events 77.3 44.5 43.5 56.2
  1. BioNLP F1 scores on the development section of the Genia track (task 1) for several event categories and overall F1. The best scores from each column are bolded.