University of Cambridge > Talks.cam > Machine Learning Reading Group @ CUED > Large Margin Training of Hidden Markov Models

Large Margin Training of Hidden Markov Models

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Shakir Mohamed.

Large Margin Training of Hidden Markov Models in Speech Recognition.

In particular, I will make first an introduction to automatic speech recognition followed by some popular approaches to train such systems (ML, MMIE ). Next I will highlight the main weaknesses of them and introduce some of the alternative training frameworks. Mainly I will focus on Large Margin Training applied to HMMs. Finally, I will give a description of the state of the art system used to transcribe broadcast news in three languages (English, Arabic and Mandarin) developed here at University of Cambridge.

The paper on the Large Margin Training:

Fei Sha, Lawrence K. Saul, “Large Margin Hidden Markov Models for Automatic Speech Recognition”, NIPS 2006 , http://books.nips.cc/papers/files/nips19/NIPS2006_0143.pdf

The corresponding PhD thesis containing sequential derivation of Large Margin Training algorithms for GMM , observable Markov Model and, finally, Hidden Markov Model:

Fei Sha, “Large Margin Training of Acoustic Models for Speech Recognition”, University of Pennsylvania, 2007. http://www-rcf.usc.edu/~feisha/pubs/thesis_tree.pdf

Short summary on Large Margin training of Gaussian Mixture Models:

Fei Sha, Lawrence K. Saul, ” Large Margin Gaussian Mixture Modeling for Phonetic Classification and Recognition”, Proc. ICASSP , 2006. http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=1660008&isnumber=34757

This talk is part of the Machine Learning Reading Group @ CUED series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

© 2006-2019 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity