EAT: Enhanced ASR-TTS for Self-supervised Speech Recognition
Baskar, Murali Karthick, Burget, Lukáš, Watanabe, Shinji, Astudillo, Ramon Fernandez, Černocký, Jan "Honza''
–arXiv.org Artificial Intelligence
Self-supervised ASR-TTS models suffer in out-of-domain data conditions. Here we propose an enhanced ASR-TTS (EAT) model that incorporates two main features: 1) The ASR$\rightarrow$TTS direction is equipped with a language model reward to penalize the ASR hypotheses before forwarding it to TTS. 2) In the TTS$\rightarrow$ASR direction, a hyper-parameter is introduced to scale the attention context from synthesized speech before sending it to ASR to handle out-of-domain data. Training strategies and the effectiveness of the EAT model are explored under out-of-domain data conditions. The results show that EAT reduces the performance gap between supervised and self-supervised training significantly by absolute 2.6\% and 2.7\% on Librispeech and BABEL respectively.
arXiv.org Artificial Intelligence
Apr-13-2021
- Country:
- Europe > Czechia > South Moravian Region > Brno (0.04)
- Genre:
- Research Report > New Finding (0.48)
- Technology: