Skip to main content

Table 10 Effect of dataset size reduction on single-task and multi-task performance

From: A neural network multi-task learning approach to biomedical named entity recognition

 

1.0

0.5

0.5

0.25

0.25

0.1

0.1

Dataset

STM

STM

MO-MTM

STM

MO-MTM

STM

MO-MTM

AnatEM

81.55

78.74*

78.35*

74.82*

76.59*+

65.99*

63.15

BC2GM

72.63

70.27*

70.73*+

67.37*

67.14*

63.07*

63.14*

BC4CHEMD

82.95

80.16*

79.22*+

76.81*

76.26*

71.94*

72.53*

BC5CDR

83.66

81.15*

82.45*+

79.09*

80.44*+

74.47*

75.48*

BioNLP09

83.90

81.89*

82.22*

80.56*

79.58*

75.12*

78.32*

BioNLP11EPI

77.72

74.00*

77.57*+

70.89*

75.61+

67.63*

75.04*+

BioNLP11ID

81.50

76.65

81.39

70.60*

78.17*+

68.19*

73.52*

BioNLP13CG

76.74

70.58*

75.02*+

65.08*

72.98*+

51.61*

67.86*+

BioNLP13GE

73.28

73.32

81.37*+

67.43

78.80*

52.66*

77.12*+

BioNLP13PC

80.61

75.39*

77.57

70.03*

73.90*

57.62*

68.65*+

CRAFT

79.55

75.25*

79.01+

72.19*

76.79*+

60.91*

71.00*

Ex-PTM

68.56

62.81

74.60*+

53.30*

74.27*+

47.01*

69.83+

JNLPBA

69.60

68.34

69.65

66.63*

68.13

62.80*

65.40*+

Linnaeus

83.98

80.08*

87.61+

69.53*

79.86

39.44

45.73

NCBI-Disease

80.26

76.51

76.84

71.88*

73.55*

67.48*

62.89*

Average

78.43

75.01

78.24

70.41

75.47

61.73

68.64

  1. (Bold: best scores for dataset, Italic: better score for each setting, *: statistically significant compared to full single-task model, +: statistically significant compared to corresponding single-task model)