Improvements on the pronunciation prefix tree search organization

Author

Alleva, Fil ; Huang, Xuedong ; Hwang, Mei-Yuh

Author_Institution

Microsoft Corp., Redmond, WA, USA

Volume

1

fYear

1996

fDate

7-10 May 1996

Firstpage

133

Abstract

The need for ever more efficient search organizations persists as the size and complexity of the knowledge sources used in continuous speech recognition (CSR) tasks continues to increase. We address efficiency issues associated with a search organization based on pronunciation prefix trees (PPTs). In particular we present (1) a mechanism that eliminates redundant computations in non-reentrant trees, (2) a comparison of two methods for distributing language model probabilities in PPTs, and (3) report results on two look ahead pruning strategies. Using the 1994 DARPA 20 k NAB word bigram for the male segment of si dev5m 92 (the 5k speaker independent development test set for the WSJ), the error rate was 12.2% with a real-time factor of 1.0 on a 120 MHz Pentium

Keywords

natural languages; probability; speech recognition; tree searching; 120 MHz; 1994 DARPA 20 k NAB word bigram; Pentium processor; WSJ; efficiency; error rate; knowledge sources; language model probabilities; look ahead pruning strategies; nonreentrant trees; pronunciation prefix tree search organization; real-time factor; speaker independent development test set; Decoding; Delay effects; Distributed computing; Error analysis; History; Lattices; Natural languages; Speech recognition; System testing; Target recognition;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 1996. ICASSP-96. Conference Proceedings., 1996 IEEE International Conference on

Conference_Location

Atlanta, GA

ISSN

1520-6149

Print_ISBN

0-7803-3192-3

Type

conf

DOI

10.1109/ICASSP.1996.540308

Filename

540308