| Presentation # | 15 |
| Session: | ASR IV |
| Location: | Kallirhoe Hall |
| Session Time: | Friday, December 21, 13:30 - 15:30 |
| Presentation Time: | Friday, December 21, 13:30 - 15:30 |
| Presentation: |
Poster
|
| Topic: |
Speech recognition and synthesis: |
| Paper Title: |
SPEAKER ADAPTATION FOR END-TO-END CTC MODELS |
| Authors: |
Ke Li, Johns Hopkins University, United States; Jinyu Li, Yong Zhao, Kshitiz Kumar, Yifan Gong, Microsoft AI and Research, United States |
| Abstract: |
We propose two approaches for speaker adaptation in end-to-end (E2E) automatic speech recognition systems. One is Kullback-Leibler divergence (KLD) regularization and the other is multi-task learning (MTL). Both approaches aim to address the data sparsity especially output target sparsity issue of speaker adaptation in E2E systems. The KLD regularization adapts a model by forcing the output distribution from the adapted model to be close to the unadapted one. The MTL utilizes a jointly trained auxiliary task to improve the performance of the main task. We investigated our approaches on E2E connectionist temporal classification (CTC) models with three different types of output units. Experiments on the Microsoft short message dictation task demonstrated that MTL outperforms KLD regularization. In particular, the MTL adaptation obtained 8.8% and 4.0% relative word error rate reductions (WERRs) for supervised and unsupervised adaptations for the word CTC model, and produced 9.6% and 3.8% relative WERRs for the mix-unit CTC model, respectively. |