Skip to main content

Advertisement

(b) Scenario 2: "adaptation of neutral ASR engine"

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

    M   E   A
Baseline system 65.0 81.0 79.2
Adapted systems    
 Acoustic models 64.5 83.1 83.6
Linguistic models 65.9 81.6 81.6
Both 65.9 84.4 85.1
  1. Levels of significance (adjusted according to [34]):
  2. 0.05   0.01   0.005   0.002   0.001.