The path to a text file that contains a list of new words to add to the language model at load time, and a list of weightings to give to specific words in the language model.
The text file must be in the following format:
1 machine 1 Nerf 0.5
where the first line specifies how many classes are in the file, the second line specifies the class word and the number of words in that class, and the subsequent lines contain a list of words and the weighting to give those words.
Note: Your class word should contextually resemble the words which are being added. Thus, in the previous example the class word machine
precedes the word gun
, and the word Nerf
also usually precedes the word gun
.
Any word listed under a class
adopts the statistical figures of the word specified as the class name, and the weighting specified for the specific words. Thus, in the previous example the word Nerf
adopts the statistical figures of the machine
class in the language model, weighted by 0.5
.
The recommended weight for words is in the range 0.25 - 0.5.
You can also specify an optional pronunciation for words in the class word file, by adding it after the weighting. For example:
Nerf 0.5 n er f
Action: | AddTask, CheckResources |
---|---|
Task: | WavToText, StreamToText, TelWavToText, StreamToTextMusicFilter, WavToTextPunct, StreamToTextPunct, TelWavToTextPunct, StreamToTextMusicFilterPunct, ClusterSpeechToTextTel |
Type: | String |
Default: | None |
Example: |
|
See Also: | ClassWordFile (configuration parameter) |
|