TY - GEN
T1 - Attention-based end-to-end speech recognition on voice search
AU - Shan, Changhao
AU - Zhang, Junbo
AU - Wang, Yujun
AU - Xie, Lei
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/9/10
Y1 - 2018/9/10
N2 - Recently, there has been a growing interest in end-to-end speech recognition that directly transcribes speech to text without any predefined alignments. In this paper, we explore the use of attention-based encoder-decoder model for Mandarin speech recognition on a voice search task. Previous attempts have shown that applying attention-based encoder-decoder to Mandarin speech recognition was quite difficult due to the logographic orthography of Mandarin, the large vocabulary and the conditional dependency of the attention model. In this paper, we use character embedding to deal with the large vocabulary. Several tricks are used for effective model training, including L2 regularization, Gaussian weight noise and frame skipping. We compare two attention mechanisms and use attention smoothing to cover long context in the attention model. Taken together, these tricks allow us to finally achieve a character error rate (CER) of 3.58% and a sentence error rate (SER) of 7.43% on the MiTV voice search dataset. While together with a trigram language model, CER and SER reach 2.81% and 5.77%, respectively.
AB - Recently, there has been a growing interest in end-to-end speech recognition that directly transcribes speech to text without any predefined alignments. In this paper, we explore the use of attention-based encoder-decoder model for Mandarin speech recognition on a voice search task. Previous attempts have shown that applying attention-based encoder-decoder to Mandarin speech recognition was quite difficult due to the logographic orthography of Mandarin, the large vocabulary and the conditional dependency of the attention model. In this paper, we use character embedding to deal with the large vocabulary. Several tricks are used for effective model training, including L2 regularization, Gaussian weight noise and frame skipping. We compare two attention mechanisms and use attention smoothing to cover long context in the attention model. Taken together, these tricks allow us to finally achieve a character error rate (CER) of 3.58% and a sentence error rate (SER) of 7.43% on the MiTV voice search dataset. While together with a trigram language model, CER and SER reach 2.81% and 5.77%, respectively.
KW - Attention model
KW - Automatic speech recognition
KW - End-to-end speech recognition
KW - Voice search
UR - http://www.scopus.com/inward/record.url?scp=85054248254&partnerID=8YFLogxK
U2 - 10.1109/ICASSP.2018.8462492
DO - 10.1109/ICASSP.2018.8462492
M3 - 会议稿件
AN - SCOPUS:85054248254
SN - 9781538646588
T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
SP - 4764
EP - 4768
BT - 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018
Y2 - 15 April 2018 through 20 April 2018
ER -