#include <julius/julius.h>
Go to the source code of this file.
This file has functions for score calculations on the 2nd pass. It includes Viterbi path update calculation of a hypothesis, calculations of scores and word trellis connection at word expansion.
The cross-word triphone will be computed just at word expansion time, for precise scoring. This is called "nextscan" altgorithm. These functions are enabled when PASS2_STRICT_IWCD is DEFINED in config.h. If undefined, the "backscan" functions in search_bestfirst_v1.c will be used instead.
Here in nextscan algorithm, all cross-word context dependencies between next word and source hypothesis are computed as soon as a new hypotheses is expanded. As the precise cross-word triphone score is applied on hypothesis generation with no delay, more accurate search-time score can be obtained than the delayed backscan method in search_bestfirst_v1.c. On the other hand, the computational cost grows much by re-calculating forward score of cross-word triphones for all the generated hypothethes, even non-promising ones.
Definition in file search_bestfirst_v2.c.