基于RNN Transducer端到端语音识别的最小贝叶斯风险训练 | 论文解读

{"type":"doc","content":[{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"一、概述"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"端到端语音识别技术将语音识别系统中的各个组件整合至同一个神经网络框架中，与传统语音识别系统相比具有建模简洁，赋能组件之间联合优化以及系统占用空间小等优点，近几年逐渐成为语音识别领域里最重要的研究方向之一。现有的端到端语音识别系统主要包括基于 Connnectionist Temporal Classification (CTC)，基于 Sequence-to-sequence(Seq2Seq) 以及基于 RNN Transducer (RNNT) 三类系统。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}