Seek Geo
Entertainment, Life Story, Travel & News
Auto-Captioning: FAIL ?
Like many of you, I was thrilled about auto-captioning on Google/YouTube so I checked with a few friends of mine like Roberto and RiverBirch who tested it.
They all said the exact same thing: “It failed.” RiverBirch tested it out, and believe it or not, the captions came out different than what she said. We are aware that it is not going to be 100% accurate, but how can it be almost all wrong?
Here’s the video of RiverBirch for an example:
Not seeing CC? Go to bottom on right corner to turn CC on.
Correct Transcription:
Hi everyone, this is River.
This is a test.
Youtube has put into effect automatic captioning for videos.
The only problem is that it takes a few days to come into effect.
So this is a test video to see if it works.
Like others, I am also floored by their interpretation. What have they done wrong? Any ideas? Accents problem? Not loud enough?
Buy me a cup of coffee
14 Responses to “Auto-Captioning: FAIL ?”
Youtube gets AUTOMATIC CC videos - AllDeaf.com — March 7, 2010 @ 8:16 pm
RSS feed for comments on this post. TrackBack URI
Leave a reply
Click on pics above to their blogs. How to join: Click here
Large Chatbox: Click here Rule: Click here
Click above to subscribe via email.
Recent Comments
Recent Posts
Categories
Syndication Options
Subscribe to this site's RSS feed.
yes, Fail!
What I did was I loaded my video to you tube then went over to Captions and Subtitles, selected ‘Machine Transcription’ and let it finish processing. I came back 20 minutes later to check on the video and it says “Machine Transcription (Failed)”.
There was another option ‘Add a Caption Track’ so I selected that and decided to do a transcript file. I typed word for word everything that was said in the video and saved it as a Word doc. I uploaded that file to the video and it finished. I reviewed the video and right away it was covered with errors. A few seconds into the video the CC started looking fine then it messed up again not even five seconds later. So phhhfffttt!! you tube just needs to work this out and see what they can do with it.
Roberto, you mean it even screwed up what you typed in? That is making me furious.
Hey, Geo!
Google said that the captioning program is not perfect yet, which is why there is the option of manually correcting the captioning when someone submits a video with speech in it to YouTube.
It’s gonna take a while for folks to catch onto the manual correction option, that is, if they ever bother to use it!!! :P
At least this is better than NO captioning, and Google is to be commended for making this effort when many other online video sources haven’t. As I’ve said in a previous blog article of mine, it’s comes across in volumes to SET a better example instead of being made an example by law. Google is setting that example, and let’s hope others will follow without having to be forced. :)
How do you manually correct?
I agree, it’s a major fail on some of them.. When I was watching a video last night, (sound turned off) I was reading what it said, and I was thinking, “Wow, this don’t seem right?” I then replayed it with sound turned on, and it was NO where close!
On how to use the manual option, click http://www.google.com/support/youtube/bin/answer.py?hl=en&answer=100077.
Ann_C
I should say that the manual option only works for those uploading videos to You Tube, BTW.
Thanks! I will look at the manual work. I’ve been using overstream.
Thanks guys for being helpful with tips!
Hi Ann_C!
Oh yea, that is true that it is not perfect. I was just puzzled as of why it is not even 1% accurate. We all knew there will be many faults but barely any correct ones?
I watched Roberto’s video (he removed captions right after) and it was like HUGE mess. Not single correct one so with RiverBirch, almost all wrong ones, too. I saw other video last night, not even a single correct one as well.
So why released the awesome feature right away knowing it’s not even working? What are we missing here? That’s the reason I brought it up.
I guess due to overhype with all about auto-captioning in the past making it sounds like it works near perfectly, he he.
You are absolutely right, this is better than NO captioning, though.
By the way, I had no idea you have a blog! I see there a link to your name *bookmarked* Time to do some catching-up!
Many thanks!!
youTube gives a few suggestions to improving the quality of auto-caps:
No background noise and speaking clearly into the mic.
Also, if she has a script she was following (doesn’t have to be time-coded), she can upload it along with the video and youTube will timecode the words for her.
Last of all, she can download the file of timecoded subtitles and fix some of the words and upload it back into youTube. This is the major achievement, in my opinion, as it removes 95% of the subtitling work and makes it much easier for everyone to subtitle their clips.
Jared
Ah.. many thanks, Jared! Now this is good for us to know. It’s not easy as it sounds but at least there is something only if anyone would be willing to do it.
Again, thanks for tips!
As a guy who does some volunteer voice captioning, I am pretty sure these issues have to do with voice diction. In order for me to get good recognition, I have to concentrate on making sure my words don’t blend together, and that I speak clearly. In the test above, it seems like it should have been more accurate, but she spoke with a bit of a lisp, that affected the accuracy.
I don’t know how they can improve, unless they start profiling each person that has to be recognized.
I’m sorry to say that I don’t anticipate the auto captioning getting much better in the near future, if ever. The main problem is that computers don’t have the ability to understand language, so they can’t discriminate between things that sound alike based on context. I think people don’t realize how often they use that ability when listening to speech, because it happens semi-subconsciously. I wrote an article on it, more specifically talking about speech versus voice recognition versus stenocaptioning/CART in college situations, but the same principles apply in the case of automatic captioning in YouTube videos.
http://stenoknight.com/VoiceVersusCART.html