from    
to    
search  

 


学习优化
第476期“工物学术论坛”:Combining mechanistic models and data-driven algorit...
Conversion of Biomass into Useful Chemicals and Biofuel
Publishing with Chem from Cell Press(腾讯会议:389-181-164)
报告题目:
『清华信息大讲堂』第 63讲:Self-Validated Story Segmentation of Chinese Broadcast News Using Subword Normalized Cuts
 报告人:
Prof. Liu,Zhi-Qiang
city university of Hong Kong
报告时间:
2010-10-26 09:30
报告地点:
信息楼(FIT)1-312
主办单位:
  简介:

Abstract:

Automatic story segmentation is an important prerequisite for semantic--‐level content retrieval. The normalized cuts (NCuts) method has recently shown great promise for segmenting English spoken lectures. However, the availability assumption of the exact story number per document significantly limits its feasibility to handle a large number of files. Besides, how to apply such method to another language in the presence of speech recognition errors is unknown yet. This talk I will addresses the two problems by proposing a self validated NCuts (SNCuts) algorithm for segmenting Chinese broadcast news via inaccurate lexical cues, generated by the Chinese large vocabulary continuous speech recognizer (LVCSR). Due to the specialty of Chinese LVCSR, we present a subword level graph embedding to the erroneous transcripts. We regularize the NCuts criterion by an exponential prior of story numbers, respecting the principle of Occam's razor. Given the maximum story number as a general parameter, our method can obtain reasonable segmentations for a large number of news transcripts, with the story numbers automatically determined for each file, and with only slightly increased complexity. Extensive experiments on benchmark datasets show that: (i)our SNCuts algorithm can efficiently produce comparable (or even better) segmentation quality, as compared to other state of the art methods; (ii)the subword embedding always contributes to recovering the lexical cohesion of Chinese erroneous transcripts, thus improving both the segmentation accuracy and robustness to LVCSR errors.

Biography:

Dr Liu has worked in both academic and industry. He has taught in three Universities in Canada, the University of Melbourne, and City University of Hong Kong, SAR, China. Since joining the School of Creative Media, he has initiated a number of research and development projects on media/graphics systems that offer media students, content creators, and users new, intelligent active platforms. He has successfully supervised numerous, mostly PhD, postgraduate students in computer science, media technologies, and more recently, media art.

Dr Liu has served numerous international conferences as chairman or member of program committees. He is an Associate Editor and Guest Editor for more a dozen major international journals. Professor Liu was the Vice-Chairman and Chairman of IEEE Systems, Man and Cybernetics (SMC) Society, Hong Kong Chapter, co-founder and Chairman of IEEE SMC International Technical Committee on Media Computing (TCMC), and a member of IEEE International Technical Committee on Fuzzy Systems (TCFS). Dr Liu was invited as distinguished Professor by universities in China, Canada and Italy and presented many keynote speeches at international conferences. He was an IEEE Distinguished Lecturer in IEEE SCMS.

Professor Liu was the Director and Principal Investigator of Computer Vision and Machine Intelligence Lab (CVMIL)and then of the Intelligent Systems Lab (ISL) at the Department of Computer Science and Software Engineering, The University of Melbourne, Australia.

As part of the China's Space Program initiatives in the late of 70s, in 1978 after going through a series of excruciatingly rigorous National Examinations and Screenings, Dr Liu was selected by the Chinese Government to join the first handful of Chinese postgraduate students to study Aerospace Science and Engineering in North America and Western Europe. He was fortunate to have studied at the world leading school, the Institute for Aerospace Studies (UTIAS), The University of Toronto, Ontario, Canada.

Dr Liu was a Senior/Principal systems engineer in electronics and communications industry in China and Canada. Ever since the mid 1970s he participated or led the following major research and development projects: China's first generation of inertial guidance systems for navigation of missiles and aircraft; Channel equalization schemes in wireless and mobile communications; Code Division Multiplex Access (CDMA) coding scheme initiatives (written in C programming language), which was supported in part by the Canadian Government; Wireless data communications, in particular, for mobile office scenarios; and Internet/Intranet wireless data communications in enterprises.

今日相关信息
海峡论坛第二十场:台湾信息(咨讯)通讯现...
清华大学外文系•喜迎百年校庆系列...
化学系迎百年校庆系列学术报告: XPS and...
重新认识东亚近代的知识体系
 
同类别相关信息
E级超级计算机研制的若干关键问题
云端AI HPC计算能力及计算环境研究
今日与未来:清华大学钱学森班首席教授与...
新时代中国经济发展之路
无人机发展及其对构建低空网络的机遇
学术活动