19 Nov 1999
Sphinx Speech Group, CMU-SCS (rkm@cs.cmu.edu)
Lextree Search: Motivation
·Most active HMMs are word-initial models, decaying rapidly subsequently
·On 60K-word Hub-4 task, 55% of active HMMs are word-initial
·(Same reason for handling left/right contexts differently.)
·But, no. of distinct word-initial model types much fewer:
·
·
·
·
·Use a “prefix-tree” structure to maximize sharing among words
START S-T-AA-R-TD STARTING S-T-AA-R-DX-IX-NG STARTED S-T-AA-R-DX-IX-DD            STARTUP S-T-AA-R-T-AX-PD START-UP S-T-AA-R-T-AX-PD