Just to comment further, auto-caption accuracy varies greatly by person and even within a person. A lot of speech input users find they need to create different profiles such as a morning and afternoon voice profile. I knew one person who even had spring versus non-spring variations due to her allergies.
Training data overwhelmingly focused on the midwestern American accent until recently. This is largely due to the historical source of the training data for a long time: newscasts with teleprompters. A lot of speech recognizers were great about recognizing Muammar Gaddafi due to this. Regional accent recognition is getting better, but it can be very stereotyped, and this isn't even getting into foreign language accents. Some are better supported, such as Indian, especially if it's one more local to Bangalore. I leave the reason for why this is to the reader. Domain-specific terms and jargon are particularly tricky. Some recognition systems now involve the source content, like the built-in captions in PowerPoint. I believe that's the only reason why the recognition spells my last name correctly. Captioning has so many nuances to it, especially with video. If I ever manage to stop putting out fires or trying to fix toxic work situations, I might get to actually write them up. Katherine (Kate) Deibel | PhD Inclusion & Accessibility Librarian Syracuse University Libraries T 315.443.7178 [email protected] 222 Waverly Ave., Syracuse, NY 13244 Syracuse University -----Original Message----- From: Code for Libraries <[email protected]> On Behalf Of Pennington, Buddy D. Sent: Thursday, December 3, 2020 11:56 AM To: [email protected] Subject: Re: [CODE4LIB] Video content creation policies? Completely agree with Kate. Auto-captioning is a decent start but you have to clean them up to make them helpful. And many users have a need for captions. So please do it. -----Original Message----- From: Code for Libraries <[email protected]> On Behalf Of Kate Deibel Sent: Thursday, December 3, 2020 10:53 AM To: [email protected] Subject: Re: [CODE4LIB] Video content creation policies? WARNING: This message has originated from an External Source. This may be a phishing expedition that can result in unauthorized access to our IT System. Please use proper judgment and caution when opening attachments, clicking links, or responding to this email. We mandate captioning as well. Kaltura does auto-captions but both myself and the university mandate checking and editing the captions soon after posting. Anyone who says the auto-captioning and pretty much perfect are made to wear the silliest of hats and sit in the corner of a circular room. Katherine (Kate) Deibel | PhD Inclusion & Accessibility Librarian Syracuse University Libraries T 315.443.7178 [email protected] 222 Waverly Ave., Syracuse, NY 13244 Syracuse University -----Original Message----- From: Code for Libraries <[email protected]> On Behalf Of Pennington, Buddy D. Sent: Thursday, December 3, 2020 11:49 AM To: [email protected] Subject: Re: [CODE4LIB] Video content creation policies? We don't have a formal policy but we are putting our library video content on Panopto and we do captioning. -----Original Message----- From: Code for Libraries <[email protected]> On Behalf Of Koshoffer, Amy (koshofae) Sent: Thursday, December 3, 2020 10:47 AM To: [email protected] Subject: Re: [CODE4LIB] Video content creation policies? WARNING: This message has originated from an External Source. This may be a phishing expedition that can result in unauthorized access to our IT System. Please use proper judgment and caution when opening attachments, clicking links, or responding to this email. Hi Kyle Would you be willing to share what you find. Here is a link the page at UC - https://www.uc.edu/about/accessibility-network/getting-started/policy.html. We use Kaltura and require captioning. And a checklist for video/audio content https://www.uc.edu/content/dam/uc/ucit/media/accessibility-network/bestpractice/best-practices-video-audio.pdf Cheers, Amy Koshoffer (pronounced "K-osh-offer") she/her(s) Assistant Director of Research & Data Services | UC Libraries [orcid small]ORCID 0000-0001-8130-103X<http://orcid.org/0000-0001-8130-103X> [email protected]<mailto:[email protected]> (preferred) | (513) 556-1310 Research & Data Services<https://libraries.uc.edu/research-teaching-support/research-data-services.html> (Data, GIS, Informatics) Cheers, Amy ________________________________ From: Code for Libraries <[email protected]> on behalf of Kyle Breneman <[email protected]> Sent: Thursday, December 3, 2020 10:32 AM To: [email protected] <[email protected]> Subject: [CODE4LIB] Video content creation policies? I'm looking for policies that govern the library's creation of public-facing video content. Anyone have an internal policy stipulating where videos are stored or mandating captioning? Kyle Breneman University of Baltimore
