中国科学院自动化研究所机构知识库

         Transcript disfluency detection (TDD) is an important com- ponent of the real-time speech translation system, which arouses more and more interests in recent years. This pa- per presents our study on adapting neural machine transla- tion (NMT) models for TDD. We propose a general training framework for adapting NMT models to TDD task rapidly. In this framework, the main structure of the model is imple- mented similar to the NMT model. Additionally, several ex- tended modules and training techniques which are indepen- dent of the NMT model are proposed to improve the perfor- mance, such as the constrained decoding, denoising autoen- coder initialization and a TDD-specific training object. With the proposed training framework, we achieve significant im- provement. However, it is too slow in decoding to be prac- tical. To build a feasible and production-ready solution for TDD, we propose a fast non-autoregressive TDD model fol- lowing the non-autoregressive NMT model emerged recently. Even we do not assume the specific architecture of the NMT model, we build our TDD model on the basis of Transformer, which is the state-of-the-art NMT model. We conduct exten- sive experiments on the publicly available set, Switchboard, and in-house Chinese set. Experimental results show that the proposed model significantly outperforms previous state-of- the-art models.