---
abstract: "In this paper, we consider a generalization of the state-of-art\r\ndiscriminative method for optimizing the conditional likelihood in\r\nHidden Markov Models (HMMs), called the Extended Baum-Welch (EBW)\r\nalgorithm, that has had significant impact on the speech recognition\r\ncommunity. We propose a generalized form of EBW update rules that\r\ncan be associated with a weighted sum of updated and initial models,\r\nand  demonstrate that using novel update rules can significantly\r\nspeed up parameter estimation for Gaussian mixtures."
altloc: []
chapter: ~
commentary: ~
commref: ~
confdates: ~
conference: ~
confloc: ~
contact_email: ~
creators_id:
  - kanevsky@us.ibm.com
  - dpovey@us.ibm.com
  - bhuvana@us.ibm.com
  - rish@us.ibm.com
  - tsainath@MIT.EDU
creators_name:
  - family: Kanevsky
    given: Dimitri
    honourific: Dr
    lineage: ''
  - family: Povey
    given: Daniel
    honourific: Dr
    lineage: ''
  - family: Ramabhadran
    given: Bhuvana
    honourific: Dr
    lineage: ''
  - family: Rish
    given: Irina
    honourific: Dr
    lineage: ''
  - family: Sainath
    given: Tara
    honourific: Dr
    lineage: ''
date: 2008-04-27
date_type: completed
datestamp: 2008-04-30 18:35:19
department: ~
dir: disk0/00/00/60/37
edit_lock_since: ~
edit_lock_until: ~
edit_lock_user: ~
editors_id: []
editors_name: []
eprint_status: archive
eprintid: 6037
fileinfo: /style/images/fileicons/application_pdf.png;/6037/1/ebw.pdf
full_text_status: public
importid: ~
institution: ~
isbn: ~
ispublished: ~
issn: ~
item_issues_comment: []
item_issues_count: 0
item_issues_description: []
item_issues_id: []
item_issues_reported_by: []
item_issues_resolved_by: []
item_issues_status: []
item_issues_timestamp: []
item_issues_type: []
keywords: 'Extended Baum Welch, optimization, speech'
lastmod: 2011-03-11 08:57:06
latitude: ~
longitude: ~
metadata_visibility: show
note: ~
number: ~
pagerange: ~
pubdom: TRUE
publication: ~
publisher: ~
refereed: FALSE
referencetext: "Cong Liu   Peng Liu   Hui Jiang   Soong, F.   Ren-Hua Wang, Constrained Line Search Optimization for Discriminative Training in Speech Recognition, Proc. ICASSP, 2007.\r\n\r\nDimitri Kanevsky, Extended Baum Transformations for General Functions, Proc. ICASSP.\r\n\r\nDimitri Kanevsky, Extended Baum Transformations for General Functions, II, tech. Rep. RC23645(W0506-120), Human Language technologies, IBM.\r\n\r\nDaniel Povey, Discriminative Training for Large Vocabulary Speech Recognition, PhD Thesis March 1, Cambridge, UK.\r\n\r\nDaniel Povey, Brian Kingsbury, Evaluation of Proposed Modifications to MPE for Large Scale Discriminative Training, ICASSP'07.\r\n\r\nDaniel Povey, Dimitri Kanevsky, Brian Kingsbury, Bhuvana Ramabhadran, George Saon and Karthik Visweswariah, Boosted MMI for model and feature-space discriminative training, ICASSP'08.\r\n\r\nTara N. Sainath, Dimitri Kanevsky, Bhuvana Ramabhadran, Gradient Steepness Metrics Using Extended Baum-Welch Transformations for Universal Pattern Recognition Tasks, Proc. ASRU, 2007\r\n\r\nP.C. Woodland and D. Povey, Large Scale Discriminative Training of hidden Markov modeles for Speech Recognition, Computer Speech and Language, pp. 25-47, Vol. 16, No. 1, January 2002.\r\n"
relation_type: []
relation_uri: []
reportno: ~
rev_number: 22
series: ~
source: ~
status_changed: 2008-04-30 18:35:19
subjects:
  - comp-sci-speech
succeeds: ~
suggestions: ~
sword_depositor: ~
sword_slug: ~
thesistype: ~
title: 'A New Family of Extended Baum-Welch Update Rules '
type: preprint
userid: 6589
volume: ~
<script src="//archive-bar.soton.ac.uk/archive-bar.js"></script>