Minimum Bayes risk training of CTC acoustic models in maximum a posteriori based decoding framework

@article{Kanda2017MinimumBR,
  title={Minimum Bayes risk training of CTC acoustic models in maximum a posteriori based decoding framework},
  author={Naoyuki Kanda and Xugang Lu and Hisashi Kawai},
  journal={2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  year={2017},
  pages={4855-4859}
}
When using connectionist temporal classification (CTC) based acoustic models (AMs) for large vocabulary continuous speech recognition (LVCSR), most previous studies have used a naive interpolation of the CTC-AM score and an additional language model score, although there is no theoretical justification for such an approach. On the other hand, we recently… CONTINUE READING

3 Figures & Tables