Combination of strongly and weakly constrained recognizers for reliable detection of OOVS

被引：22

作者：

Burget, Lukas ^{[1
]}

Schwarz, Petr ^{[1
]}

Matejka, Pavel ^{[1
]}

Hannemann, Mirko ^{[2
]}

Rastrow, Ariya ^{[3
]}

White, Christopher ^{[3
]}

Khudanpur, Sanjeev ^{[3
]}

Hermansky, Hynek ^{[1
,4
,5
]}

Cernocky, Jan ^{[1
]}

机构：

[1] Brno Univ Technol, Speech FIT, Brno, Czech Republic

[2] Univ Magdeburg, Magdeburg, Germany

[3] Johns Hopkins Univ, Baltimore, MD 21218 USA

[4] IDIAP Res Inst, Martigny, Martigny, Switzerland

[5] Ecole Polytech Fed Lausanne, Lausanne, Switzerland

来源：

2008 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING, VOLS 1-12 | 2008年

基金：

美国国家科学基金会;

关键词：

LVCSR; OOV; confidence measures;

D O I：

10.1109/ICASSP.2008.4518551

中图分类号：

O42 [声学];

学科分类号：

070206 ; 082403 ;

摘要：

This paper addresses the detection of OOV segments in the output of a large vocabulary continuous speech recognition (LVCSR) system. First, standard confidence measures from frame-based word-and phone- posteriors are investigated. Substantial improvement is obtained when posteriors from two systems - strongly constrained (LVCSR) and weakly constrained (phone posterior estimator) are combined. We show that this approach is also suitable for detection of general recognition errors. All results are presented on WSJ task with reduced recognition vocabulary.

引用

页码：4081 / +

页数：2

共 12 条

[1]

[Anonymous], P IEEE ICASSP

[2]

Bazzi I, 2002, Modelling out-of-vocabulary words for robust speech recognition

[3]

BERNARDIS G, 1998, P ICSLP 98 SYDN AUST

[4]

BISANI M, P INTERSPEECH 2005

[5]

FOUSEK P, 2008, P ICASSP 2008 LAS VE

[6]

HOFFMEISTER B, 2007, P ICASSP 2007 HAW

[7]

KETABDAR H, 2007, P INTERSPEECH 2007 A

[8]

LJOLJE A, 1999, P EUROSPEECH 99 BUD, P99

[9]

MANGU L, 1999, P EUROSPEEC 99 BUD

[10]

SCHWARZ P, 2006, P ICASSP 2006 TOUL

← 1 2 →