A Survey on Conditional Random Fields in Automatic Speech Recognition

PhD Qualifying Examination


Title: "A Survey on Conditional Random Fields in Automatic Speech Recognition"

by

Mr. Dongpeng Chen


Abstract:

As an acoustic model, hidden Markov model (HMM) has dominated the field for 
more than 30 years for its power to model temporal speech sequences and 
computational efficiency. However, the first-order Markov chain and the 
conditional independence assumptions of HMM, which are made to simplify the 
computation, also limit the modeling power. In recent years, various 
alternative models are proposed with the purpose to beat HMM in either 
recognition accuracy or computational cost.

Conditional random felds (CRF), a framework for building probabilistic models 
to segment and label sequence data, offers several advantages over hidden 
Markov models, including the ability to relax strong independence assumptions 
made in HMM. It has been proved successful in Natural Language Processing, and 
also on tasks of computer vision.

By reviewing the conventional application of HMM in the fields of ASR, we aim 
to learn the experience and lessons from the past. More importantly, we will 
compare the HMM and CRF frameworks on ASR. Finally, several applications of CRF 
in ASR will be introduced.


Date:                   Monday, 9 January 2012

Time:                   2:30pm - 4:30pm

Venue:                  Room 3501
                         lifts 25/26

Committee Members:	Dr. Brian Mak (Supervisor)
                         Prof. Siu-Wing Cheng (Chairperson)
 			Dr. Raymond Wong
 			Prof. Nevin Zhang


**** ALL are Welcome ****