Minimum Bayes Risk Training for End-to-End Speaker-Attributed ASR.
Naoyuki KandaZhong MengLiang LuYashesh GaurXiaofei WangZhuo ChenTakuya YoshiokaPublished in: CoRR (2020)
Keyphrases
- end to end
- bayes risk
- automatic speech recognition
- speech recognition
- loss function
- upper and lower bounds
- congestion control
- posterior probability
- reproducing kernel hilbert space
- training set
- lower bound
- supervised learning
- multiscale
- regularization parameter
- squared error
- vector quantizer
- special case
- multiresolution
- pairwise
- distortion measure