Skip to main content

Table 2 Results for the gene-disease corpus.

From: Extraction of semantic biomedical relations from text using conditional random fields

 

Recall

Precision

F-score

Dictionary + naive rule-based

43.31

42.98

43.10

CRF + naive rule-based

67.62

71.88

69.68

one-step CRF

73.36

78.66

75.90

cascaded CRF

76.61

79.46

78.00

CRF + SVM

76.63

79.48

78.03

  1. NER and SRE performance comparison of one-step and cascaded CRF with three benchmark methods. The cascaded CRF is on par with the CRF+SVM model, where the latter one requires an expensive preceding feature selection step.