Speech Recognition on LibriSpeech 960 clean (test)
2.27WERProposed Method
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Proposed MethodBias list size (N)=100, Biasing weight (lambda)=4.42025.12 | 2.27 | 8.7 | 2.07 | |
| Proposed MethodBias list size (N)=200, Biasing weight (lambda)=4.42025.12 | 2.28 | 9.02 | 2.07 | |
| Proposed MethodBias list size (N)=500, Biasing weight (lambda)=4.42025.12 | 2.3 | 9.27 | 2.08 | |
| Proposed MethodBias list size (N)=100, Biasing weight (lambda)=1.02025.12 | 2.34 | 10.98 | 2.07 | |
| Proposed MethodBias list size (N)=200, Biasing weight (lambda)=1.02025.12 | 2.36 | 11.24 | 2.08 | |
| Proposed MethodBias list size (N)=500, Biasing weight (lambda)=1.02025.12 | 2.37 | 11.49 | 2.09 | |
| AED-MTPBias list size (N)=02025.12 | 2.58 | 17.27 | 2.27 | |
| AED-MTPBias list size (N)=1002025.12 | 2.58 | 17.27 | 2.27 | |
| AED-MTPBias list size (N)=2002025.12 | 2.58 | 17.27 | 2.27 | |
| AED-MTPBias list size (N)=5002025.12 | 2.58 | 17.27 | 2.27 | |
| Proposed MethodBias list size (N)=0, Biasing weight (lambda)=1.02025.12 | 2.61 | 17.78 | 2.13 | |
| Proposed MethodBias list size (N)=0, Biasing weight (lambda)=4.42025.12 | 2.61 | 17.78 | 2.13 | |
| Baseline (AED)Bias list size (N)=02025.12 | 2.73 | 17.52 | 2.27 | |
| CLASBias list size (N)=2002025.12 | 3.08 | 15.11 | 2.71 | |
| CLASBias list size (N)=1002025.12 | 3.09 | 15.56 | 2.7 | |
| CLASBias list size (N)=02025.12 | 3.12 | 15.56 | 2.73 | |
| CLASBias list size (N)=5002025.12 | 3.13 | 15.75 | 2.74 | |
| MELDSize=260M, Hrs=960, Beam search size=52026.05 | 3.5 | — | — | |
| dMel (ASR)Size=258M, Hrs=960, Beam search size=52026.05 | 4.2 | — | — | |
| MELDSize=200M, Hrs=960, Beam search size=52026.05 | 4.2 | — | — | |
| MELD w/o SpecAugSize=200M, Hrs=960, Beam search size=52026.05 | 4.5 | — | — | |
| MoshiSize=7B, Hrs=7M, Beam search size=52026.05 | 5.8 | — | — | |
| Codec-LMSize=200M, Hrs=960, Beam search size=52026.05 | 6.4 | — | — | |
| Codec-LM w/o codebook initSize=200M, Hrs=960, Beam search size=52026.05 | 100 | — | — |