In boundary-mode language identification, IDOL Speech Server initially segments audio by language and then identifies the language in each segment. Short segments can lower identification accuracy and you can choose to exclude short segments from language identification by setting DiscardShort
to a minimum segment length.
Tasks return a result of 'Unknown
' for any section with a duration below the specified value.
Action: | AddTask
CheckResources |
Task: | LangId |
Type: | Decimal |
Default: | 5.0 |
Example: | DiscardShort=1.0
|
See Also: | DiscardShortSegs (configuration parameter) |
|