Abstract | ||
---|---|---|
We explore training attention-based encoder-decoder ASR in low-resource settings. These models perform poorly when trained on small amounts of transcribed speech, in part because they depend on having sufficient target-side text to train the attention and decoder networks. In this paper we address this shortcoming by pretraining our network parameters using only text-based data and transcribed speech from other languages. We analyze the relative contributions of both sources of data. Across 3 test languages, our text-based approach resulted in a 20% average relative improvement over a text-based augmentation technique without pretraining. Using transcribed speech from nearby languages gives a further 20-30% relative reduction in character error rate. |
Year | DOI | Venue |
---|---|---|
2019 | 10.21437/Interspeech.2019-3254 | INTERSPEECH |
DocType | Citations | PageRank |
Conference | 0 | 0.34 |
References | Authors | |
0 | 6 |
Name | Order | Citations | PageRank |
---|---|---|---|
Matthew Wiesner | 1 | 5 | 2.85 |
Adithya Renduchintala | 2 | 1 | 1.74 |
Shinji Watanabe | 3 | 1158 | 139.38 |
Chunxi Liu | 4 | 23 | 3.28 |
N. Dehak | 5 | 1269 | 92.64 |
Sanjeev Khudanpur | 6 | 2155 | 202.00 |