Name
Abstract | ||
|---|---|---|
In this paper, we report state-of-the-art results on LibriSpeech among end-to-end speech recognition models without any external training data. Our model, Jasper, uses only 1D convolutions, batch normalization, ReLU, dropout, and residual connections. To improve training, we further introduce a new layer-wise optimizer called NovoGrad. Through experiments, we demonstrate that the proposed deep architecture performs as well or better than more complex choices. Our deepest Jasper variant uses 54 convolutional layers. With this architecture, we achieve 2.95% WER using beam-search decoder with an external neural language model and 3.86% WER with a greedy decoder on LibriSpeech test-clean. We also report competitive results on the Wall Street Journal and the Hub5u002700 conversational evaluation datasets. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.21437/interspeech.2019-1819 | Conference of the International Speech Communication Association |
DocType | Volume | Citations |
Journal | abs/1904.03288 | 1 |
PageRank | References | Authors |
0.36 | 0 | 8 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Jason H. Li | 1 | 153 | 15.18 |
| Vitaly Lavrukhin | 2 | 4 | 1.78 |
| Ginsburg, Boris | 3 | 75 | 8.77 |
| Ryan Leary | 4 | 11 | 1.55 |
| Oleksii Kuchaiev | 5 | 224 | 10.94 |
| Jonathan M. Cohen | 6 | 236 | 20.25 |
| huyen nguyen | 7 | 11 | 9.77 |
| Ravi Teja Gadde | 8 | 2 | 1.40 |