Closed captioning for YouTube…possibilities for Livestream? (5 posts)

|
Share Topic
  • Profile picture of SaraA saraa5p said 3 months, 1 week ago:

    Apparently you can get an automatic transcription for captioning on YouTube, which someone can then go in and clean up. This means that hearing impaired folks can follow it, and that people can get translations. I was unable to determine how to do this for Livestream, but am hoping that someone will be able to point us in the right direction.

    Of course, I didn’t learn this from anybody in the anti-racism committee, including the supposed advocates for accessibility; I had to learn it at an FCC meeting. This demonstrates why I left the anti-racism committee. Nonetheless, this forum is where people are likely to look so I’m hoping someone will the technical knowledge of how to close-caption Livestream will see it here.

    Once we have it set up, we will need volunteers to clean up the transcriptions.

    | #
  • Profile picture of Jose jose said 3 months, 1 week ago:

    On YouTube, the viewer controls whether automated transcriptions or translations based on those transcriptions are displayed. In the lower right edge of the video there is the resolution setting (240p, 360p etc.) and next to it is a CC button (closed captioning) that appears when you start playing the video. For the video you are about to view, you enable viewing with the automatic transcription or a machine translation into another language of it through the menu accessed by clicking on that CC button. By clicking on the word “settings” in either the menu that comes up from the CC button or the one accessed by clicking on the resolution, you go to a setting page where you can set the transcripts to be shown by default.

    The quality of the transcription is imperfect: it uses the same technology as Google Voice’s transcription of messages. The translations are very imperfect because errors in the transcription mushroom in the translation. Typically, translation software works not by substituting one word at a time, for often there are several possible choices. So the software takes into account surrounding words to simulate what a reader does by understanding the context. So a mistaken word in the transcription affects not only the rendering of that word in the target language, but also possibly other words.

    Commercial services that offer real-time closed captioning will also do it for streamed distribution just like they do for cable or over-the-air television.

    Neither livestream nor ustream offer real-time transcripts, not even the automated machine-only transcripts offered on YouTube. You can get programs that do this sort of automated transcribing, and I guess one could put together a setup where video would be streamed to our server, and there the closed captioning data would be generated and added, perhaps by altering the image to “burn in” the caption text.

    | #
  • Profile picture of Aimi aimi1p said 3 months, 1 week ago:

    Sara, I’d like to kindly remind you that you only attended one of our committee meetings, so I’m not sure how it would be possible for you to get a fair assessment of which technologies were discussed. I’d also like to remind you that the “supposed advocates for disability” you are talking about are people with disabilities who have direct experience with different technologies and have tried to be helpful. There’s no need to exercise privilege to disrespect anyone’s knowledge or question their commitment. We’re all trying to figure out how to make it possible for us to participate, but accessibility is fostered through empowerment, not by speaking and doing for others or making folks with experience with inaccessibility feel like they are being a burden. So let’s try to avoid the personal attacks in public forums and just brainstorm solutions, okay? :) We wouldn’t expect any less self-reflexivity in our anti-racist or feminist work. Hopefully we’ll all be more inclined to want to participate that way.

    There is unfortunately no easy solution, but with a little troubleshooting something can at least get underway and be adjusted as needed. I agree with Jose that in my experience, Youtube/Google Voice are not options that tend toward accuracy or the greatest amount of accessibility. If you have Google Voice, you may have received some rather funny voice mail transcriptions.

    However, if that solution is being proposed due to resource limitations, there are a few things to try depending on capacity:

    -To test whether the Youtube option will work, upload a video of a past GA and see how it transcribes speakers, the people’s mic, and random noise from multiple people talking at once. That way, you should theoretically be able to tell what will need to be edited out of the transcription (if garbled noise is transcribed). Maybe try one outside and one inside video to determine the range of transcription capacity. Alternatively, videos of GAs from other cities on Youtube could be checked.
    -A computer programmer may be able to create a program that automatically deletes repeated words and phrases from transcriptions (in the event that the people’s mic is used).
    -An audio engineer may be able to make a basic edit to the audio track to take out noise. Alternatively, if there is a way to get whomever is speaking to do it directly into a recording microphone, the transcription will be more accurate (about as accurate as Google Voice).
    -Some folks do voice transcription for accessibility on a volunteer basis for things like online films. Someone may be willing to do this for GA’s after the fact or while watching the livestream. If there is a way to project this during meetings, it could work and does not require folks to be physically present if they are not able to be. Occupy Deaf and Decolonize Disability are starting to set up networks of folks who do this in different cities. OA could similarly recruit dedicated transcribers who are sympathetic to the movement.
    -SIRI on iphone does some transcription. If there is a dedicated phone that can be used for this purpose (maybe it can be the mic people speak into) that might work but I’m unsure of transcription quality.

    The ATL folks whose contact info I forwarded before who do accessible interfaces. You have indicated that there are people working on access already. Perhaps in the time since our last discussion, they have tried these suggestions and could give you an answer about whether Youtube will be a good option.

    | #
  • Profile picture of SaraA saraa5p said 3 months, 1 week ago:

    YouTube’s transcription is *possible*, we already have YouTube channels and can take volunteers to clean up the transcriptions and it will cost no money. Those may be its sole virtues, but they are important ones. If anybody has a better idea AND the resources to implement it….AND the time and energy to make it happen….then you are free to do that, or recruit people to do it. If you know people willing to do transcription who are already practiced at it, they should talk to Rob Call as he has access to Livestream and YouTube.

    | #
  • Profile picture of Jose jose said 3 months, 1 week ago:

    I am not sure why I should invest time on this as opposed to other things. The transcriptions are there, albeit imperfect, for those that need that resource, but especially given that, I don’t see it worth a lot of time perfecting them in the case of my videos. I realize others might have the time or give this greater priority but I believe it would have to be me for right now, as the YouTube channel that I’ve posted on is tied to my Google accounts that I wouldn’t feel comfortable handing out the login/passwords to. Probably poor planning on my part, but the specific login is tied to gmail, google voice, etc, through browser cookies. After I finish a big project I’m working on, I’ll spend some time trying to figure out how to reconfigure things so the credentials to work on the videos can be shared or I put the on a (new) channel not tied to my personal things.

    | #