Skip to main content

Table 3 Precision/Recall/F1-score results for gene mention detection over CRAFT development set: ABNER with distributed model trained on BioCreative I using different evaluation mapping strategies

From: A corpus of full-text journal articles is a robust evaluation tool for revealing differences in performance of biomedical natural language processing tools

 

ABNER BioCreative

protein-STAR

ABNER BioCreative

protein-GENE

ABNER BioCreative

protein-POLYSTAR

 

Prec

Recall

F1

Prec

Recall

F1

Prec

Recall

F1

strict

0.35

0.46

0.40

0.12

0.31

0.18

0.20

0.62

0.30

overlap

0.50

0.69

0.58

0.23

0.64

0.34

0.23

0.74

0.35

shared

0.49

0.65

0.56

0.22

0.57

0.32

0.23

0.73

0.35

subspan

0.50

0.69

0.58

0.23

0.64

0.34

0.23

0.74

0.35