Through Google Cloud Speech-to-Text and Natural Language API, StoryCorps is excited to offer machine-generated transcripts for the interviews in the StoryCorps Online Archive. We understand that this will raise many questions for our users and participants. You can find the answers to many common questions below:
Are all interviews transcribed?
Not yet, but they will be! When we launched this new feature in November of 2020, the majority of English-language Signature interviews (interviews recorded with a StoryCorps facilitator or StoryKit) were transcribed. Transcripts for new interviews are currently being added tot he archive in batches and interviews recorded with the StoryCorps App and Connect, and non-English interviews will be transcribed in the near future. You can learn more here.
What is a machine-generated transcript?
We use the term “machine-generated transcript” to indicate that these transcripts are created through technology rather than by a person. While this technology makes transcribing the large number of interviews in the Online Archive possible, machine-generated transcripts may contain inaccuracies, misspellings, or omissions.
Can other people see my transcript?
That depends on your privacy/visibility setting. Anyone who can access your interview, based on your privacy setting, can also access the transcript. You can learn more about your privacy/visibility settings here.
Who can download my transcript?
Currently, transcripts cannot be downloaded from archive.storycorps.org. We look forward to making it possible for participants in the interview to download their transcripts in the near future.
What if I see an error in my transcript?
Because of the extremely large number of transcripts newly available in the Online Archive, StoryCorps’ ability to correct transcripts is limited at this time. We intend to begin offering to make corrections to transcripts on a limited basis in the near future.
If there is an error in your transcript that you consider important to be corrected, please reach out StoryCorps Public Support. You can learn more here.
What if my transcript is in the wrong language?
Currently, StoryCorps is only able to offer transcripts of English-language interviews. We look forward to offering this in the coming months. If your interview is not in English but was transcribed, there may have been a technical error. Please reach out to StoryCorps Public Support by submitting a request
Why aren’t the speakers’ names indicated in the transcript?
“Speaker diarization” is the technical term for the feature that would allow a machine-generated transcript to accurately indicate which speaker is talking. Our transcription technology is not yet able to provide speaker diarization with the level of accuracy we need to use this feature. We look forward to adding speaker diarization in the near future.
If you have concerns about speaker diarization in your interview transcript for purposes of accessibility, please reach out to StoryCorps Public Support by submitting a request.
How does the addition of these transcripts affect the privacy of interview participants?
We understand that for our participants, the scope of this project can bring up many questions about their personal information, their interview, and their privacy. You can find details around many privacy questions related to transcripts here.