Skip to main content

Table 7 Scenario 2: adaptation of the acoustic and linguistic models; results are given in terms of word accuracy (%). The baseline system (acoustic and linguistic models trained on Ohm_base) is given in column "Ohm_base" and is identical in all three tables. "" denotes the arithmetic (unweighted) mean. The average of the four subsets weighted by the prior probabilities of the four classes is given in row "Mont."

From: On the Impact of Children's Emotional Speech on Acoustic and Language Models

 

Acoustic models trained on

 

Ohm_base

Ohm_base +

Ohm_base +

Ohm_base +

Ohm_base +

Test set

(baseline)

2x Ohm_M

1x Ohm_N

3x Ohm_E

2x Ohm_A

Mont_M

65.0

64.5

61.5

59.9

61.3

Mont_N

77.3

77.6

77.5

77.1

78.0

Mont_E

81.0

81.3

80.5

83.1

81.2

Mont_A

79.2

80.2

78.8

81.4

83.6

75.6

75.9

74.6

75.4

76.0

Mont

77.5

77.7

77.5

77.4

78.2

 

Linguistic models trained on

 

Ohm_base

Ohm_base +

Ohm_base +

Ohm_base +

Ohm_base +

Test set

(baseline)

28x Ohm_M

1x Ohm_N

28x Ohm_E

28x Ohm_A

Mont_M

65.0

65.9

64.5

64.0

64.5

Mont_N

77.3

77.0

77.4

77.7

77.7

Mont_E

81.0

80.1

80.8

81.6

81.9

Mont_A

79.2

78.9

79.0

79.9

81.6

75.6

75.5

75.4

75.8

76.4

Mont

77.5

77.1

77.5

77.8

78.0

 

Acoustic models trained on

 

Ohm_base

Ohm_base +

Ohm_base +

Ohm_base +

Ohm_base +

 

(baseline)

0x Ohm_M

1x Ohm_N

3x Ohm_E

2x Ohm_A

 

Linguistic models trained on

 

Ohm_base

Ohm_base +

Ohm_base +

Ohm_base +

Ohm_base +

Test set

(baseline)

28x Ohm_M

1x Ohm_N

28x Ohm_E

28x Ohm_A

Mont_M

65.0

65.9

61.5

60.4

59.1

Mont_N

77.3

77.0

77.6

77.4

78.4

Mont_E

81.0

80.1

80.4

84.4

83.1

Mont_A

79.2

78.9

78.7

81.6

85.1

75.6

75.5

74.6

76.0

76.4

Mont

77.5

77.1

77.6

77.8

78.7