Transcribing broadcast news with the 1997 Abbot System
Acoustics, Speech and Signal Processing, 1998. Proceedings of the 1998 IEEE International Conference（1998）
摘要
Recent DARPA CSR evaluations have focused on the transcrip- tion of broadcast news from both television and radio programmes (17). This is a challenging task because the data includes a variety of speaking styles and channel conditions. This paper describes the development of a connectionist-hidden Markov model (HMM) system, and the enhancements designed to improve performance on broadcast news data. Both multilayer perceptron (MLP) and recurrent neural network acoustic models have been investigated. We asses the effect of using gender-dependent acoustic models, and the impact on performance of varying both the number of pa- rameters and the amount of training data used for acoustic mod- elling. The use of context-dependent phone models is described, and the effect of the number of context classes is investigat ed. We also describe a method for incorporating syllable boundary infor- mation during search. Results are reported on the 1997 DARPA Hub-4 development test set.
更多查看译文
关键词
acoustic signal processing,broadcasting,hidden Markov models,learning (artificial intelligence),multilayer perceptrons,recurrent neural nets,speech recognition,telecommunication computing,1997 Abbot System,DARPA CSR evaluations,DARPA Hub-4 development test set,HMM system,MLP,acoustic models,broadcast news data,broadcast news transcription,channel conditions,connectionist-hidden Markov model,context-dependent phone models,gender-dependent acoustic models,multilayer perceptron,radio programmes,recurrent neural network,speaking styles,speech recognition system,syllable boundary information,television programmes,training data
AI 理解论文
溯源树
样例
生成溯源树，研究论文发展脉络
Chat Paper
正在生成论文摘要