Efficient 2-Pass N-Best Decoder

Efficient 2-Pass N-Best Decoder (1997)

Cached

Download Links

by Long Nguyen , Richard Schwartz , Fawcett St

Venue:	DARPA Speech Recognition Workshop
Citations:	8 - 5 self

Versions

Version History

Metadata Version 3

User correction supplied by SystemCorrections

Datum	Value	Source
TITLE	Efficient 2-Pass N-Best Decoder	INFERENCE
AUTHOR NAME	Long Nguyen	user correction - Legacy Corrections
AUTHOR AFFIL	BBN Systems & Technologies	user correction - Legacy Corrections
AUTHOR NAME	Richard Schwartz	user correction - Legacy Corrections
AUTHOR AFFIL	BBN Systems & Technologies	user correction - Legacy Corrections
AUTHOR NAME	Fawcett St	user correction - Legacy Corrections
AUTHOR ADDR	Cambridge, MA. 02138	user correction - Legacy Corrections
ABSTRACT	In this paper, we describe the new BBN BYBLOS efficient 2-Pass N-Best decoder used for the 1996 Hub-4 Benchmark Tests. The decoder uses a quick fastmatch to determine the likely word endings. Then in the second pass, it performs a time-synchronous beam search using a detailed continuousdensity HMM and a trigram language model to decide the word starting positions. From these word starts, the decoder, without looking at the input speech, constructs a trigram word lattice, and generates the top N likely hypotheses. This new 2-pass N-Best decoder maintains comparable recognition performance as the old 4-pass N-Best decoder, while its search strategy is simpler and much more efficient. 1. INTRODUCTION As previously described in [2], the old BBN BYBLOS decoder used a multi-pass search strategy consisting of 4 passes to generate the top N most likely hypotheses, which were then rescored using more detailed, but expensive knowledge sources. These N best hypotheses were then reordered and th...	user correction - Legacy Corrections
YEAR	1997	INFERENCE
VENUE	DARPA Speech Recognition Workshop	INFERENCE
VENUE TYPE	CONFERENCE	INFERENCE
PAGES	167--170	INFERENCE
CITATIONS	4 found	ParsCit 1.0