seq2seq_tds

Sequence-to-Sequence Speech Recognition with Time-Depth Separable Convolutions (Hannun et al., 2019)

The architecture and flags files are provided to reproduce results from the paper (training and decoding steps). Considered benchmarks:

• Librispeech

Citation

@article{DBLP:journals/corr/abs-1904-02619,
  author    = {Awni Hannun and
               Ann Lee and
               Qiantong Xu and
               Ronan Collobert},
  title     = {Sequence-to-Sequence Speech Recognition with Time-Depth Separable
               Convolutions},
  journal   = {CoRR},
  volume    = {abs/1904.02619},
  year      = {2019},
  url       = {http://arxiv.org/abs/1904.02619},
  archivePrefix = {arXiv},
  eprint    = {1904.02619},
  timestamp = {Wed, 24 Apr 2019 12:21:25 +0200},
  biburl    = {https://dblp.org/rec/bib/journals/corr/abs-1904-02619},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}