Improved Lexical Tree Search For Large Vocabulary Speech Recognition

Improved Lexical Tree Search For Large Vocabulary Speech Recognition (1998)

Cached

Download Links

[www-i6.informatik.rwth-aachen.de]

by Stefan Ortmanns , Andreas Eiden , Hermann Ney

Venue:	In Proceedings of the ICASSP
Citations:	9 - 3 self

Versions

Version History

Metadata Version 3

User correction supplied by SystemCorrections

Datum	Value	Source
TITLE	Improved Lexical Tree Search For Large Vocabulary Speech Recognition	user correction - Legacy Corrections
AUTHOR NAME	Stefan Ortmanns	user correction - Legacy Corrections
AUTHOR AFFIL	Lehrstuhl fur Informatik VI, RWTH Aachen-- University of Technology	user correction - Legacy Corrections
AUTHOR ADDR	52056 Aachen, Germany	user correction - Legacy Corrections
AUTHOR NAME	Andreas Eiden	user correction - Legacy Corrections
AUTHOR AFFIL	Lehrstuhl fur Informatik VI, RWTH Aachen-- University of Technology	user correction - Legacy Corrections
AUTHOR ADDR	52056 Aachen, Germany	user correction - Legacy Corrections
AUTHOR NAME	Hermann Ney	user correction - Legacy Corrections
AUTHOR AFFIL	Lehrstuhl fur Informatik VI, RWTH Aachen-- University of Technology	user correction - Legacy Corrections
AUTHOR ADDR	52056 Aachen, Germany	user correction - Legacy Corrections
ABSTRACT	This paper describes some extensions to the language model (LM) look-ahead pruning approach which is integrated into the time-synchronous beam search algorithm. The search algorithm is based on a lexical prefix tree in combination with a wordconditioned dynamic search space organization for handling trigram language models in a one-pass strategy. In particular, we study several LM look-ahead pruning techniques. Further, we improve the efficiency of this look-ahead technique by exploiting subtree dominance. This method avoids the computation of redundant subtrees within the copies of the lexical prefix tree and thus reduces the memory requirements of the search algorithm. In addition, we present a pruning criterion depending on the state index. The experimental results on the 20 000-word NAB'94 task (ARPA North American Business Corpus) indicate that the computational effort can be reduced to 4 times real time on a ALPHA5000 PC without a significant loss in the recognition accuracy.	user correction - Legacy Corrections
YEAR	1998	INFERENCE
VENUE	In Proceedings of the ICASSP	INFERENCE
VENUE TYPE	CONFERENCE	INFERENCE
PAGES	817--820	INFERENCE
VOLUME	2	INFERENCE
CITATIONS	8 found	ParsCit 1.0