Skip to main content

Table 2 Experimentation results for recognizing spatial attribute of events based on statistical machine learning approach

From: Analysis of syntactic and semantic features for fine-grained event-spatial understanding in outbreak news reports

Features

Machine learning techniques

Event class

  

Normal

Reporting

Information

Hypothetical

Over all

11 location-related features

CRF

84.8

(85.8,83.8)

82.7

(81.9,83.5)

62.6

(63.0,62.2)

88.1

(92.5,84.1)

81.3

(81.9,80.7)

 

SVM

84.6

(85.8,83.5)

83.7

(82.7,84.6)

54.8

(55.2,54.5)

81.0

(85.0,77.3)

80.0

(80.7,79.4)

 

C4.5

78.8

(79.6,77.9)

85.7

(85.2,86.2)

44.3

(45.0,43.6)

64.4

(65.1,63.6)

74.9

(75.5,74.4)

11 location-related features + event class

CRF

87.2

(88.3,86.1)

86.1

(85.6,86.6)

68.2

(68.4,67.9)

76.2

(80.0,72.7)

83.8

(84.5,83.1)

 

SVM

86.6

(87.7,85.5)

83.7

(82.7,84.6)

65.8

(66.2,65.4)

78.6

(82.5,75.0)

82.6

(83.3,82.0)

 

C4.5

82.8

(84.1,81.6)

88.4

(87.9,88.9)

55.3

(56.8,53.8)

78.2

(79.1,77.3)

80.1

(81.0,79.2)

11 location-related features + subject type

CRF

85.7

(87.4,84.1)

85.8

(84.9,86.6)

75.5

(76.0,75.0)

80.95

(85.0,77.3)

84.1

(85.1,83.1)

 

SVM

84.8

(86.4,83.2)

83.7

(82.7,84.6)

58.7

(59.1,58.3)

83.3

(87.5,79.5)

80.8

(81.6,79.9)

 

C4.5

79.5

(80.8,78.3)

87.2

(86.3,88.0)

61.3

(61.1,61.5)

61.2

(61.9,60.5)

78.0

(78.5,77.4)

11 location-related features + subject type + event class

CRF

86.7

(88.1,85.4)

87.1

(86.4,87.8)

80.4

(80.6,80.1)

76.2

(80.0,72.7)

85.5

(86.3,84.7)

 

SVM

86.1

(87.8,84.4)

84.0

(83.1,84.4)

68.4

(68.8,67.9)

81.0

(85.0,77.3)

82.8

(83.7,82.0)

 

C4.5

80.2

(80.9,79.5)

88.0

(87.5,88.5)

66.7

(68.7,64.7)

64.4

(65.1,63.6)

79.5

(80.1,78.8)

  1. The attribute annotation results shown in this table were based on micro-averaging. The scores are shown in the form of "F-score (precision, recall)"