`
`3
`
`2:33mm
`
`Nice“©920$
`
`2:399
`
`3209
`
`rmi
`
`Nr
`
`zc<
`
`mEmmzm
`
`LEE
`
`xcmEmEn—
`
`wuwtNV
`
`BEEENIV
`
`:25?
`
`Emum.:or
`
`mENc
`
`m.».o_,..88v
`
`835mmowvu
`
`
`
`
`
`
`
`
`
`
`w8cmucmw«8z
`
`
`
`20:55.8:33.
`
`
`
`‘3:.32me
`
`Emmwooi
`
`
`
`was6.550
`
`0925cm.255.
`
`mmmzmcg
`
`2/6
`
`was33?2:63,5"III
`
`8:32
`
`.202,
`
`8:33
`
`389
`
`om
`
`
`
`mzn.mucElmmwamcfl
`
`Homwmooimm.1..
`
`.338.—=mPr.
`
`3622
`
`.ommmooi
`
`boEaE
`
`Ea;
`
`£58550
`
`mm
`
`5N
`
`3.68£3
`
`m2858
`
`magi“:
`
`£033
`
`5:05:00
`
`San.mEN.moi
`
`5929:
`
`
`
`mEmcwnohmmw
`
`_‘N
`
`NNk
`
`_...—_‘
`
`x53305..an:
`32x2
`n.3335
`3:23:8.
`
`3:32:2
`
`3.338..
`
`35.338—
`
`N53305
`
`
`
`on::8me
`
`
`
`
`NE
`
`
`
`
`
`v5630Lommmooimwc..8..me
`
`
`
`
`
`boEos.Eves.camaou<
`
`v83303
`
`_.________
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`mmwmk.20.ij
`
`.rOm
`
`
`
`vmmm....20.qu
`
`#0..
`
`3/6
`
`
`
`30>H
`
`x+>>:->
`
`c+>>3->
`
`mmmi
`
`
`
`
`4/6'
`
`81
`
`INITIALISE
`
`
`
`
`DATA FOR A NEW LBCICAL
`TREE RECEIVED FROM SEARCH
`CONTROLLER?
`
`
`
`YES
`
`
`
`,
`
`35
`
`READ FEATURE VECTOR FROM
`BUFFER
`
`I
`
`EVALUATE STATE TRANSITIONS FOR
`FIRST LEXICAL TREE NODE USING
`ACOUSTIC MODEL DATA
`
`36
`'
`
`S7
`
`DETERMINE SCORE FOR FIRST
`LEXICAL TREE NODE
`
`38
`
`
`
`
`REQUEST LANGUAGE MODELSCORE
`USING PREVIOUS N-1 WORDS IN'
`RECEIVED DATA AND LEGCALTREE
`DATA IN ACOUSTIC MODEL MEMORY
`
`
`
`
`
`
`RECEIVE LANGUAGE MODEL SCORES ‘
`39 .FOR WORDS IN THE LEXICALTREE
`AND PICK HIGHEST SCORE
`
`
`
`
`
` GENERATE TEMPORARY LEXICAL TREE
`
`310 SCORE USING THE SOORE FOR FIRST
`
`
`LEXICALTREE NODE AND THE
`HIGHEST LANGUAGE MODEL SCORE
`
`
`
`811'
`
`SEND SCORETO RESULTS MEMORY
`AS TEMPORARY LBGCAL TREE SCORE
`
`Fig 4
`
`
`
`5/6
`
`
`
`820' 821
`
`
`DATA FOR A LEXTOA‘L
`TREE RECENED FROM SEARCH
`CONTROLLER?
`
`
`
`FEATURE VECTOR
`V-AILABLE IN BUFFE
`
`
` READ FEAT'EVE" OR ' 'OM
`
`BUFFER
`
`823
`
`N°
`
`ERROR
`
`
`
`EVALUATE-STATE TRANSITIONS FOR
`
`
`EACH PA’IH USING ACOUSTIC MODEL
`
`
`DATA
`
`
`DETERMINE SCORES FOR PATHS, SEND
`
`826
`BEST SCORE TO RESULTS MEMORY AND
`
`
`
`
`I STORE PATH HISTORIES LOCALLY
`
` PRUNING APPLlED T0 LEXICAL TREE
`
`TO DELETE PATHS
`
`
`828
`
`
`'ATH(S) REACHED WORD END
`
`
`
`
`
`
`
`
`
`
`YES
`
`APPLY LANGUAGE MODEL SCORE
`
`
`
`
`
`
`SEND SCORE AND HISTORY TO
`RESULTS MEMORY
`
`DELETE PATH(S) AND
`HISTORY DATA
`
`832
`
`
`
`ANY PATHS LEFT TO PROCESS?
`
`NO
`
`833 MESSAGESENTTO SEARCH CONTROLLERTO
`~
`INDICATE THAT LBGCAL TREE HAS‘BEEN
`PROCESSED
`
`Fig 5
`
`
`
`6/6
`
`840 INITIALISATION
`
`s41
`
`READ INITIAL LFJGGALTREE DATA IN
`RESULTS MEMORY
`
`S42
`
`INITIAL LEGCAL TREE DATA
`
`
`DISTRIBUTED AMONGST LBGCAL‘TREE
`
`
`PROCESSORS FOR TEMPORARY
`
`
`LEXICAL TREE SCORE DETERMINATION
`
`
`
`.
`
`‘2
`'
`s43 TEMPORARY LEXICALTREE SCORES
`RErURNED TO RESULTS MEMORY
`
`344
`
`PRUNE THE LEXICAL TREES IN THE
`RESULTS MEMORY ON BASIS OF
`TEMPORARY LEXICAL TREE SCORES
`
`
`LEXICAL TREE PROCESSING
`S45 DISTRIBUTED AMONGST LEXICAL TREE
`
`PROCESSORS
`
`
`S46
`
`
`‘ ISTORY AND 800 '
`
`
`ENTERED IN RESULTS MEMORY FOR
`
`
`
`
`
`
`'
`1"
`F Ig 0
`
`
`DETERMINE NEXT POSSIBLE LEXICAL
`
`
`TREES USING CROSS WORD
`
` 847‘
`TRIPHONES
`
`
`
`848
`
`
`
`S49
`
`350
`
`PRUNE THE LEXICAL TREES IN "IT-IE
`
`RESULTS MEMORY ON BASIS OF
`
`TEMPORAR¥ LEXICAL SCORES
`
`
`I
`
`
`
`
`
`352
`
`RESULTS
`OUTPUT
`
`LEXICAL TREE DATA DISTRIBUTED
`
`AMONGST LEXICAL TREE
`ALL
`
`
`PROCESSORS FOR TEMPORARY
`LEXICAL TREE
`
`LEXICAL‘TREE SCORE DETERMINATION
`PROCESSORS FINISHED
`
`
`PROCESSING AND NO LEXICAL
`
`
`' TREES IN RESULTS
`
`
`TEMPORARY LEXICAL TREE SCORES 7
`RETURNED TO RESULTS MEMORY
`
`
`
`