AUTOMATED TRANSCRIPTION
Faster Whisper Embedded
Automated Transcription – Faster Whisper Embedded
Transana integrates server-based speech recognition technology from Faster Whisper.
This option processes all data on your computer, without involving any external computers. This is especially important for sensitive human subjects data, where an IRB might not permit transmission and processing of your data on an external server.
Automated Transcription – Overview
When creating a new Transcript or editing Transcript Properties, you have the option of selecting an Automated Transcription Method on the Transcript Properties form. (Please note that if you have a transcript in this item, it will be replaced by the automated transcript.)
At this time, you can select Faster Whisper Embedded, Speechmatics Server or Deepgram Server.
When you have selected an Automated Transcription Method and pressed OK, Transana will run the Automated Transcription tool. This tool guides you through the process of audio extraction and submission of this audio data to the selected Automated Transcription service.
All of the services that Transana currently supports provide time information with their data, so Transana includes time codes at the start of every sentence. They also includes estimates of transcription accuracy with their data. Transana color-codes these results, marking parts of the transcription with less than 90% confidence.
Please note that automated transcripts still require manual review and correction. Automated Transcription speeds the process of transcription in many instances, but the results can be uneven at times.
Faster Whisper Embedded
If you select “Faster Whisper Embedded” from the list of options, your request for automated transcription will be processed on your own computer using the Faster Whisper tool. This tool requires that you download some data the first time you use each accuracy “model,” but your data never leaves your computer.
Select the language spoken in the media file.
Select the automated transcription “model” which determines the level of accuracy of the process, recognizing that greater accuracy requires more time in the automated transcription process.
Select the frequency of time code placement. As a general rule, adding time codes every sentence is adequate, but there are times when more frequent time coding is helpful.
If you are transcribing media in a language other than English, Faster Whisper gives you an option to create an English translation of your data. If you want both a transcript and a translation, you will need to create two separate transcripts in Transana and run the process twice.
