Knowledge Commons of Institute of Automation,CAS
Speech-Transformer: A No-Recurrence Sequence-to-Sequence Model for Speech Recognition | |
Dong, Linhao1,2; Xu, Shuang1; Xu, Bo1 | |
2018-04 | |
会议名称 | International Conference on Acoustics, Speech and Signal Processing (ICASSP) |
页码 | 5884-5888 |
会议日期 | 2018-04 |
会议地点 | Calgary, Canada |
出版者 | IEEE Xplore |
产权排序 | 1 |
摘要 | Recurrent sequence-to-sequence models using encoder-decoder architecture have made great progress in speech recognition task. However, they suffer from the drawback of slow training speed because the internal recurrence limits the training parallelization. In this paper, we present the Speech-Transformer, a no-recurrence sequence-to-sequence model entirely relies on attention mechanisms to learn the positional dependencies, which can be trained faster with more efficiency. We also propose a 2D-Attention mechanism, which can jointly attend to the time and frequency axes of the 2-dimensional speech inputs, thus providing more expressive representations for the Speech-Transformer. Evaluated on the Wall Street Journal (WSJ) speech recognition dataset, our best model achieves competitive word error rate (WER) of 10.9%, while the whole training process only takes 1.2 days on 1 GPU, significantly faster than the published results of recurrent sequence-to-sequence models. |
关键词 | speech recognition sequence-to-sequence attention transformer |
学科门类 | 工学 |
收录类别 | EI |
资助项目 | Beijing Science and Technology Program[Z171100002217015] ; Beijing Science and Technology Program[Z171100002217015] |
语种 | 英语 |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/39274 |
专题 | 复杂系统认知与决策实验室_听觉模型与认知计算 |
作者单位 | 1.Institute of Automation, Chinese Academy of Sciences, China 2.University of Chinese Academy of Sciences, China |
第一作者单位 | 中国科学院自动化研究所 |
推荐引用方式 GB/T 7714 | Dong, Linhao,Xu, Shuang,Xu, Bo. Speech-Transformer: A No-Recurrence Sequence-to-Sequence Model for Speech Recognition[C]:IEEE Xplore,2018:5884-5888. |
条目包含的文件 | 下载所有文件 | |||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
Speech-Transformer_A(640KB) | 会议论文 | 开放获取 | CC BY-NC-SA | 浏览 下载 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论