H90-1003 we present a new sentence-level fast match scheme for continuous speech
H89-2028 Methods for performing acoustic fast matches are currently known . NLP fast
H91-1055 system includes a tree-structured fast match \ -LSB- l , 3 \ -RSB- to reduce
H90-1003 computation . The second algorithm is a fast match scheme for continuous speech
H89-2028 matches are currently known . NLP fast matches may or may not be possible .
H92-1081 . THE FAST MATCH The acoustic fast match ( AFM ) algorithm used here is
H92-1067 approach depends on having a good fast match strategy to identify candidate
H89-2028 both the CSR and NLP will provide fast matches . These matches take a partial
C00-2152 index that allows a preliminary fast match . Examples that survive the fast
H91-1055 probabilities are ignored . The current fast match reduces the number of possible
H89-2028 combination of the probabilities . ( " Fast matches " can he used to limit the number
H90-1032 could be incorporated into the fast match algorithm . The seventh and last
H90-1003 Several papers have described fast match schemes that look ahead ( incurring
H92-1081 primarily on a time basis . THE FAST MATCH The acoustic fast match ( AFM
H92-1081 topic of particular interest is fast match techniques and structures . There
H89-2027 alternatives must be generated using a fast match <title> Advisory Committee :
D09-1110 the forest , result in a very fast matching procedure , as illustrated in
H92-1077 search strategy with an acoustic fast match . Cross-word models had not yet
H92-1081 decoder \ -LSB- 1,7,16 \ -RSB- with fast match \ -LSB- 2,3,5 \ -RSB- is used
H89-2028 the overall performance . 2.4 Fast matches To reduce the search space ,
hide detail