Skip to main content

Table 5 Results of the style modification process in cases of evaluation done using the trained CNN-RNN ASR model

From: Accent modification for speech recognition of non-native speakers using neural style transfer

  WM A ST
CER 46.3% 36.1% 31.7%
RICER 22% 32%
WER 56.8% 43.2% 34.9%
MSS − 0.94 1.43 1.97
CER (100-sentences) 42.1% 33.2% 28.3%
RICER (100-sentences) 21% 24.7%
WER (100-sentences) 51.8% 39.1% 32.3%
MSS (100-sentences) − 0.85 1.56 2.09