Skip to main content

Table 4 Comparison to previously published results using CRAFT-SA test data with predicted segmentation

From: Dependency parsing of biomedical text with BERT

Method Metric
LAS MLAS BLEX
Baseline 56.68 44.22 0.0
SpaCy 69.32 0.0 54.80
TurkuNLP-ST 89.70 85.55 86.63
Ours 91.21 86.63 87.68
  1. Maximum score in bold