The name of an output text file that contains far more of the word hypotheses considered during recognition, not simply the best scoring hypothesis.
Each entry in the lattice file consists of the start time, the end time, the word, and the acoustic score, with one word on each line. For example:
1754 1757 dog 2962 1758 1762 barked 1645
If a word is considered during recognition but ultimately discarded, the recall rate might be improved by looking for occurrences of the word in the lattice.
Type: | String |
Default: | None |
Required: | No |
Configuration Section: | stt module |
Example: | LatFile=audio.lat
|
See Also: | LatScale
LatWordFile |
|