Skip to main content
EURASIP Journal on Audio, Speech, and Music Processing

Table 4 A comparison of using different bias factor. WERs and CERs are in percent and evaluated on the manually segmented dev dataset

From: Timestamp-aligning and keyword-biasing end-to-end ASR front-end for a KWS system

bias factor δ

Model 1: base + FT

Model 2: FT all epochs

 

WER / CER

WER / CER

0 (baseline)

9.2/10.5

9.2/10.5

1/4

9.0/10.2

9.2/10.4

1/3

9.0/10.2

9.2/10.5

1/2

9.0/10.2

9.2/ 10.5

1

9.1/10.4

9.2/10.4

2

9.0/10.3

9.1/10.3

4

9.1/10.3

10.1/11.4

Back to article page