By default, the returned score file contains the following sections:
Matching details | Each word in the truth transcript is shown alongside the word from the speech-to-text transcript that it is matched to (truth transcript is on the left; speech-to-text transcript is on the right) and a score that shows the distance between the two words. A score of 0 (zero) indicates that the words match exactly. |
Truth analysis | For each word in the truth transcript, this section shows whether the word was correctly recognized (1) or not (0). |
Recognition analysis | For each word in the speech-to-text transcript, this section shows whether the word was correctly recognized (1) or not (0). |
Analysis details | This section shows line-by-line matching of the truth transcript with the speech-to-text transcript. |
Summary | Displays the recall, precision, and f-measure scores. |
To return the Summary section only, you can either:
ShowAlignment
and DetailedScore
configuration parameters to False
in the align
module.ShowAlignment
and ShowDetails
action parameters with the AddTask
action; the action parameters override the settings in the configuration file.For more information about these parameters, see the HPE IDOL Speech Server Reference.
|