Improvement in Automatic Speech Recognition of South Asian Accent Using Transfer Learning of DeepSpeech2

Table 8

The comparison between nonoptimized and optimized models with respect to WER and CER.

Model	Nonoptimized		Optimized
	WER	CER	WER	CER
DeepSpeech2	15.57	4.52	—	—
DeepSpeech2 KENLMc	10.46	3.68	10.45	2.96
DeepSpeech2 KENLMo	10.75	3.79	10.66	2.89
DeepSpeech2 and KENLM (c + O)	9.9	3.61	9.91	2.80