Skip to main content

Table 1 Results from WSD system. Results from WSD system applied to various sections of the NLM-WSD data set using a variety of features and machine learning algorithms. The best results obtained by our system are highlighted in bold font. Results from baseline and previously published approaches are included for comparison.

From: Disambiguation of biomedical text using diverse sources of information

 

Features

Data sets

Linguistic

CUI

MeSH

CUI+MeSH

Linguistic+MeSH

Linguistic+CUI

Linguistic+MeSH+CUI

 

Vector space model

All words

87.0

85.8

81.9

86.9

87.9

87.3

87.5

Joshi subset

82.1

79.6

76.6

81.4

83.3

82.4

82.8

Leroy subset

77.5

74.4

70.4

75.8

79.7

78.7

78.9

Liu subset

84.0

81.3

78.3

83.4

84.8

83.9

84.2

Common subset

79.1

75.1

70.4

76.9

81.1

80.0

79.7

 

Naive Bayes

All words

86.4

81.2

85.7

81.1

86.4

81.7

81.8

Joshi subset

80.9

73.4

80.1

73.7

81.1

74.1

74.5

Leroy subset

76.9

66.1

74.6

65.9

77.5

66.5

67.2

Liu subset

82.1

75.4

81.7

75.3

82.7

76.3

76.6

Common subset

77.2

66.1

74.7

65.8

79.0

66.7

67.4

 

Support Vector Machine

All words

85.9

83.5

85.3

84.5

86.2

85.3

86.0

Joshi subset

80.1

76.4

79.5

78.0

80.9

79.1

80.3

Leroy subset

75.5

69.7

72.6

72.0

77.1

74.5

76.3

Liu subset

81.7

78.2

81.0

80.0

82.3

80.6

81.7

Common subset

76.3

69.8

71.6

73.0

78.1

75.1

76.9

  

Previous Approaches

  

Per-term

Global

 

MFS baseline

Liu et al. (2004)

Joshi et al. (2005)

Leroy and Rindflesch (2005)

Joshi et al. (2005)

McInnes et. al. (2007)

All words

78.0

-

-

-

86.2

85.3

Joshi subset

66.9

-

82.5

-

80.9

80.0

Leroy subset

55.3

-

77.4

65.5

75.7

74.5

Liu subset

69.9

78.0

84.9

-

83.3

81.9

Common subset

54.9

-

79.8

68.8

78.1

75.6