YouTube’s trying out a new way to improve the accuracy of its automated captions, by letting users recommend updates and revisions based on the captions displayed.
As per YouTube:
“We’re experimenting with suggested edits to automatic captions for people watching certain EN-language YouTube videos on desktop devices. To suggest corrections, open the transcript, click the ‘pencil’ icon, and enter suggestions before clicking the ‘check mark’ icon.”
As you can see here, viewers will be able to provide edits of potential errors in the automated translation. You can then submit your updates by tapping on the tick (or checkmark) icon in the edit box.
The last icon here is ‘re-play segment’ making it easier to re-listen to a section to check again on what was said.
Any edits that you make are then sent through to YouTube – though they won’t be visible to the creator.
“Suggested edits will only be surfaced to the person who edited the transcript themselves and won’t be visible to the creator or the broader YouTube audience. We’re using this experiment to collect data on the quality of suggestions to determine how we can use suggested edits to improve the automatic captions as well as our spam and abuse protections.”
So it’s crowd-sourced editing, which is somewhat similar to the downvote option being tested on TikTok and even Twitter’s Birdwatch program, gathering feedback from users to then build into the broader content feedback and assessment loop, in the hopes of using the wisdom of the crowd to help improve each process.
Which could work. I mean, it’s better to have more eyes on a problem than relying on internal or creator review, and it could be a simple, effective way to help improve YouTube’s automated caption system, by identifying common words that trip up the system, or making it easier for YouTube to note where errors are more likely to occur, in order to alert creators.
Or maybe, if it’s really effective (and not misused), it could end up being a direct channel for viewers to alert creators to errors. There are risks in that approach as well, but maybe, if YouTube can normalize usage of the option, and it provides good feedback, that could be an expansion in future.
The test is only being conducted on ‘a small percentage of videos’ via the desktop app, so you may not run into it. But it could be the start of an expanded process to improve YouTube’s automated captioning systems.
