Paper ID | SPE-10.2 | ||
Paper Title | NON-AUTOREGRESSIVE TRANSFORMER ASR WITH CTC-ENHANCED DECODER INPUT | ||
Authors | Xingchen Song, Zhiyong Wu, Tsinghua University, China; Yiheng Huang, Chao Weng, Dan Su, Tencent, China; Helen Meng, Chinese University of Hong Kong, China | ||
Session | SPE-10: Speech Recognition 4: Transformer Models 2 | ||
Location | Gather.Town | ||
Session Time: | Tuesday, 08 June, 16:30 - 17:15 | ||
Presentation Time: | Tuesday, 08 June, 16:30 - 17:15 | ||
Presentation | Poster | ||
Topic | Speech Processing: [SPE-GASR] General Topics in Speech Recognition | ||
IEEE Xplore Open Preview | Click here to view in IEEE Xplore | ||
Abstract | Non-autoregressive (NAR) transformer models have achieved significantly inference speedup but at the cost of inferior accuracy compared to autoregressive (AR) models in automatic speech recognition (ASR). Most of the NAR transformers take a fixed-length sequence filled with MASK tokens or a redundant sequence copied from encoder states as decoder input, they cannot provide efficient target-side information thus leading to accuracy degradation. To address this problem, we propose a CTC-enhanced NAR transformer, which generates target sequence by refining predictions of the CTC module. Experimental results show that our method outperforms all previous NAR counterparts and achieves 50x faster decoding speed than a strong AR baseline with only 0.0 ~ 0.3 absolute CER degradation on Aishell-1 and Aishell-2 datasets. |