作者其他论文
文献详情
A deep neural network approach for sentence boundary detection in broadcast news
文献类型:会议
作者:Xu, Chenglin[1]  Xie, Lei[2]  Huang, Guangpu[3]  Xiao, Xiong[4]  Chng, Eng Siong[5]  Li, Haizhou[6]  
机构:[1]Shaanxi Provincial Key Laboratory of Speech and Image Information Processing, School of Computer Science, Northwestern Polytechnical University, China |Temasek Laboratories NTU, Nanyang Technological University, Singapore, Singapore
[2]Shaanxi Provincial Key Laboratory of Speech and Image Information Processing, School of Computer Science, Northwestern Polytechnical University, China
[3]Temasek Laboratories NTU, Nanyang Technological University, Singapore, Singapore
[4]Temasek Laboratories NTU, Nanyang Technological University, Singapore, Singapore
[5]Temasek Laboratories NTU, Nanyang Technological University, Singapore, Singapore |School of Computer Engineering, Nanyang Technological University, Singapore, Singapore
[6]Temasek Laboratories NTU, Nanyang Technological University, Singapore, Singapore |School of Computer Engineering, Nanyang Technological University, Singapore, Singapore |Institute for Infocomm Research, ASTAR, Singapore, Singapore
年:2014
通讯作者:Xu, Chenglin
会议名称:15th Annual Conference of the International Speech Communication Association: Celebrating the Diversity of Spoken Languages, INTERSPEECH 2014
页码范围:2887-2891
会议地点:Singapore, Singapore
会议开始日期:2014-09-14
会议结束日期:2014-09-18
收录情况:EI(20144600198898)  
所属部门:计算机学院
人气指数:868
浏览次数:851
语言:外文
摘要:
This paper presents a deep neural network (DNN) approach to sentence boundary detection in broadcast news. We extract prosodic and lexical features at each inter-word position in the transcripts and learn a sequential classifier to label these positions as either boundary or non-boundary. This work is realized by a hybrid DNN-CRF (conditional random field) architecture. The DNN accepts prosodic feature inputs and non-linearly maps them into boundary/non-boundary posterior probability outputs. Su ...More
0
评论(0 条评论)
登录