Language Modeling on Mandarin Voice Assistant (VA) (test)
0.051WERAA M-FOFE
Evaluation results
| Method | Links | |
|---|---|---|
| AA M-FOFEApplication Strategy=Application-Agnostic, Model Architecture=Mixture FOFE, Number of Parameters=34M2023.05 | 0.051 | |
| AD FOFEApplication Strategy=Application-Dependent, Model Architecture=FOFE, Number of Parameters=31M2023.05 | 0.0512 | |
| AD FOFEApplication Strategy=Application-Dependent, Model Architecture=FOFE, Number of Parameters=34M2023.05 | 0.0512 | |
| AA M-FOFEApplication Strategy=Application-Agnostic, Model Architecture=Mixture FOFE, Number of Parameters=37M2023.05 | 0.0513 | |
| AS FOFEApplication Strategy=Application-Specific, Model Architecture=FOFE, Number of Parameters=68M2023.05 | 0.0514 | |
| AS FOFEApplication Strategy=Application-Specific, Model Architecture=FOFE, Number of Parameters=58M2023.05 | 0.0517 | |
| AA FOFEApplication Strategy=Application-Agnostic, Model Architecture=FOFE, Number of Parameters=29M2023.05 | 0.0525 | |
| AA FOFEApplication Strategy=Application-Agnostic, Model Architecture=FOFE, Number of Parameters=34M2023.05 | 0.0526 | |
| AA Transf.Application Strategy=Application-Agnostic, Model Architecture=Transformer, Number of Parameters=27M2023.05 | 0.0888 | |
| AA Transf.Application Strategy=Application-Agnostic, Model Architecture=Transformer, Number of Parameters=34M2023.05 | 0.0903 |