Show simple item record

dc.contributor.authorRenals, Steve
dc.contributor.authorHochberg, Mike
dc.date.accessioned2006-06-22T10:02:24Z
dc.date.available2006-06-22T10:02:24Z
dc.date.issued1995
dc.identifier.citationIn Proc IEEE ICASSP, pages 596-599, Detroit, 1995.en
dc.identifier.urihttp://hdl.handle.net/1842/1270
dc.description.abstractIn this paper we present a novel, efficient search strategy for large vocabulary continuous speech recognition (LVCSR). The search algorithm, based on stack decoding, uses posterior phone probability estimates to substantially increase its efficiency with minimal effect on accuracy. In particular, the search space is dramatically reduced by phone deactivation pruning where phones with a small local posterior probability are deactivated. This approach is particularly well-suited to hybrid connectionist/hidden Markov model systems because posterior phone probabilities are directly computed by the acoustic model. On large vocabulary tasks, using a trigram language model, this increased the search speed by an order of magnitude, with 2% or less relative search error. Results from a hybrid system are presented using the Wall Street Journal LVCSR database for a 20,000 word task using a backed-off trigram language model. For this task, our single-pass decoder took around 15 times realtime on an HP735 workstation. At the cost of 7% relative search error, decoding time can be speeded up to approximately realtime.en
dc.format.extent19329 bytes
dc.format.mimetypeapplication/octet-stream
dc.language.isoen
dc.publisherIEEEen
dc.subjectlarge vocabulary continuous speech recognitionen
dc.subjectspeech recognitionen
dc.subjectMarkov modelen
dc.titleEfficient search using posterior phone probability estimates.en
dc.typeConference Paperen


Files in this item

This item appears in the following Collection(s)

Show simple item record