Skip to main content

Table 3 Coreference resolution performance comparison by the average F1 value

From: Distinguished representation of identical mentions in bio-entity coreference resolution

 

Dev-F1

Test-F1

Feature-based Rule-based

Hybrid

Neural

BioNLP

     

[13]

62.4

60.9

 

 

[14]

68.6

/

 

 

[11]

63.9

48.1

  

[12]

67.5

/

  

[8]

72.2

62.0

 

[9]

63.4

51.2

  

[10]

65.6

69.5

  

CRAFT

     

[7]

45.5

46.4

  

[9]

33.9

36.0

  

[10]

/

57.0

  

  1. All the models are evaluated on the platforms provided by the task organizers