Now that you have a normalized transcript and a recognition transcript, you can check the transcript for any large, mismatched sections, and also generate some location pointers for the subsequent alignment task.
To check a transcript
Send an AddTask
action to HPE IDOL Speech Server, and set the following parameters:
Type
|
The task name. Set to TranscriptCheck . |
TxtFile
|
The normalized transcript file. |
CtmFile
|
The speech-to-text transcript file produced from the audio file. |
Out
|
The file to write a transcript that contains the approximate timestamps to. |
DiagFile
|
The alignment diagnostics file to generate. |
For example:
http://localhost:13000/action=AddTask&Type=TranscriptCheck&TxtFile=C:\data\transcript.txt&CtmFile=C:\misc\speechtext.ctm&Out=AlignResults.ctm&DiagFile=myResults.diag
This action uses port 13000
to instruct HPE IDOL Speech Server, which is located on the local machine, to compare the transcript.txt
and speechtext.ctm
files to produce the transcript file with approximate timestamps (AlignResults.ctm
), and the diagnostics file (myresults.diag
). The diagnostics file contains information on how well the speech-to-text transcript file and the normalized transcript file match.
The AlignResults.ctm
output file is in the following format:
She 0.000 255.190 was 0.000 255.190 the 0.000 255.190 guardian 0.000 255.190 of 0.000 255.190
The file contains approximate timings of the transcript words in the audio file. You can use this information as the first step to aid the subsequent alignment task.
This action returns a token. You can use the token to:
|