YouTube Transcribe Audio Fail Thread

[Trekker4747]

It seems some YouTube videos now have a closed captioning feature where it'll transcribe the audio for you. (To be fair it does point out the feature is in Beta and not likely to be accurate.) So let us pick through and find the transcriptions that are not so good!

From a minute into this VIDEO (I got to it through one of today's Cracked articles.)

The woman says: "Hi, I'm Bonnie Franklin and in a very short time you are going to be able to tap-dance. I started tap-dancing at nine years of age on the Colgate Comedy Hour with the great Donald O'Connor and I started teaching tap at the age of twelve."

YouTube's Transcription: "I'm Bonnie Franklin and in a very short time you are going to be able to tap-dance. Eyesight advancing it nine years debate on the Colgate Comedy Hour with the great Donald O'Connor and he started teaching tactic twelve."

From the same Cracked article: LINK

Latoya Jackson says: "Hi. I'm Latoya Jackson and I'd like to share my exciting new step-workout with you. I really feel being fit is critical to success in the entertainment industry. It takes hard work and dedication to stay that way and I realize when it comes to fitness working hard is essential. So take it from me Progressive Step Workout will lean you, define you and redesign your body. It will really get you pumping heart and soul."

YouTube's Transcription (trying hard to work through Latoya's Cindy Brady-esque lisp): "Tires was hijacked can make analyst at Berkley Berkley. Really care about being that is critical to success any detained industry it takes hard work and dedication to state that. I realize that when it comes to fitness working hard is essential to take it for me. Progressive Step for capital emu define you redesign your body told me any picture puppy hardened shell."

 

[M'Sharak]

I saw a clip within the last few weeks which made imaginative use of the feature (and of its inaccuracy); I can't recall now where I saw it posted, but maybe someone will recognize it from the description.

Two guys wrote a short sketch about a confrontation between two co-workers over something trivial, shot a performance on video and uploaded it to YouTube, which then generated a transcript.

Next, they rewrote the sketch to reflect what YouTube's transcription feature had interpreted the dialogue to be, shot a performance of that and uploaded it to YouTube, which then generated a transcript.

Once more, they rewrote the dialogue to match the new transcript, shot a third performance on video and uploaded it to YouTube. The clip I watched included all three versions, so that you could see how it had evolved throughout the process.

Edit:

I misremembered a couple of the details, but the process is about the same as I described:

[yt]http://www.youtube.com/watch?v=23H8IdaS3tk[/yt]

 

[Trekker4747]

It's like playing the "phone game" with someone who really sucks at listening.

It's actually pretty amusing. To it's defense transcribing what Latoya was saying in the second linked video there was hard enough for me to do, but what it comes up with was actually pretty funny.

Sure it's still in Beta so it's certainly going to have kinks but what it comes up with sometimes is just awesome.

 

[Owain Taggart]

They've had that feature now for over a year, actually, but I'm guessing it's been scaled back some because it doesn't work so well, that and I think that people have to enable it when uploading videos to have it encoded. That said, it can make for some really entertaining results.

 

[Nerys Ghemor]

This is a lot like the DS9 episode "Babel." I have to hand it to the actors for being able to memorize and put the "right" emotions into the nonsensical scripts!

 

[Kommander]

This is the single one that has happened. If it were to be a castle and lost in the attic then it would be a long time before one can come of age. I really like how soon the bicycle appears in the full moon. It's like a controlled act of espionage that failed to occur. Sometimes it's like no one can achieve that which must be understood before it's too late to become that which is destined to be denied. I must concur that certain outcomes are difficult to maintain before that which is unmade can be redistributed.

 

[RJDiogenes]

I think Spambots use this feature before they post.

 

[RJDiogenes]

I misremembered a couple of the details, but the process is about the same as I described:

That was fantastic. "Kinky."

 

[tafkats]

I watched the entire run of the '90s sitcom "Cybill" on YouTube a while back and wrote down some of the better transcription fails. In one case, the same line said three times in a row -- "shoulda kept barking" (it made sense in context) -- was rendered as "shikha barking," "they can't buy in," and "ticket backing."

Here are a few more:

MARYANNE: Can you believe that all of a sympathy card? "And spiny." Right here in Memphis f. t. mother's funeral? She brought the sickness anti-Castro world.
CYBILL: She brought the suit a diploma outstripped a casserole.
MARYANNE: That was a sick person can sell? I thought was a six a two putting uh.
CYBILL: Three integrating the? So-called by governments trying to get a permanent. But I think that was it a day.
MARYANNE: That was a sweet it's identical? I thought it was a -- captured this game.

CYBILL: Congratulations on four years of glorious Dick listening.
MARYANNE: All participants have! Cruise ship rescuer coast cerebral? Sampling pictures that fat people playing shuffleboard. I'm overwhelmed.
CYBILL: I needed is the exterior shots on the commercial undoing and.
MARYANNE: Sucks folks! Think yahoo!
CYBILL: Cacao Ayatollah point was my best friend when Hong Kong treaty.

(What's funny is that "fat people playing shuffleboard" is one of the phrases it got completely right.)

 

[tafkats]

Can you name these Trek episodes as parsed by YouTube?

MILES: And the messenger from the queen watch the little man as he does stand sang and left, how he would soon have the daughter he always wanted because they would never guess that he's been mmos --
MOLLY: Want to lump will.
MILES: Will still scan, right. And the next day, the command came and passed now wishes quaking blockage by name. She said at first, "Uh ... are you called track." And no, Leon said. "But are you called a battery." Go and we had too. And what she said, "Then perhaps you and me is ..."
MOLLY: Rumbles caring.
MILES: And of the sound of his own name d old man was so furious studies stamped his fortune came the entire kingdom any broken to disappeared. The king McQueen, her daughter, lived happily ever after. (Pause.) Could make a good one. (Kisses Molly's forehead.) Street drains.
MOLLY: Yeah.
KEIKO: Bombing extension. (Kisses Molly's forehead.) The tax return.
(Miles and Keiko leave.)
MILES: To attack you into the bedtime story too.
KEIKO: Warm Olympic again Whitney.
MILES: Lecturing.

O'BRIEN: K go on these past few days its final destination another one living in those waters. I want to set up a little workshop in the bedroom --
BASHIR: You set up a workshop in the bedroom.
O'BRIEN: Used much specific gruesome she's on trying to live like a bachelor again. Expressing subconscious desire push for a out about Walters outback.
BASHIR: These ridiculous.
O'BRIEN: So I said.
BASHIR: In anything by spending your free time in the bedroom accuplacer intimately associated with kanka you up actually expressing he decided not to be closer to her with during that since unit. Stock option really.
O'BRIEN: Exactly exactly. The warmest and quite catchy CDA white catching the more likely --
BASHIR: And --
O'BRIEN: The welcome mat.
BASHIR: Stone review and wish got was a map.

GEORDI: Had another idea that might help us to talk and personable Scandinavian.
DATA: Would be happy to.
GEORDI: Done some you're trying to teach the cat to use the consul.
DATA: The discharge for training program. And teaching sparked a jump down from the desk on my comment.
GEORDI: Has its own.
DATA: It's not entirely successful. Feelin supplement number two twenty one.

JANEWAY: And everything to do is concerned.
SEVEN: Now with the forces.
JANEWAY: Stand corrected. Interview.
SEVEN: Sticking to someone else.
JANEWAY: Sarajevo countries on the other side of the country legend.

SULU: Results Mozilla does investing.
SCOTTY: Transcriber result. But this jury wait awhile as long he's got to come on down camp.
SULU: Fifteen hundred miles in closing.
SPOCK: Transporter operational but just.
KIRK: Better be able to listen.

WORF: Insofar conditions by the court is a must remarks chambers.
PICARD: That's on the way there is a warning.
WORF: Tries to kill security really station one we are ready to miss transportable.
ALIEN ON COMM: Technology prices we will order that she loves it.
WORF: Front door.
PICARD: Editor as well.

WESLEY: Which you've seen. How like act when we do.
RIKER: And dot need your help you step over here and.
GUINAN: Some simple enough.
RIKER: First words at GMail but most importantly will start with something like this. You were the most beautiful woman and the galaxy. But that might not work.
GUINAN: Yes, just at work.
RIKER: And I want to point to tell you.
GUINAN: Peter afraid past me.
RIKER: Above support mike com.
WESLEY: Commander.
RIKER: Or that you might think I was online.
GUINAN: p_b_h_g select someone.

TUVOK: Finding a British satellites in orbit. They're not functional.
JANEWAY: It has taken a little closest against anesthetics.
NEELIX: Cancun. This used to be a major trading spot.
JANEWAY: Alana.
NEELIX: I'm not sure exactly what I can ...
KIM: See in signature indicates that a major solar into retirement two years ago.
NEELIX: Damaged in the nineteen years moralist.

 

[M'Sharak]

MILES: And the messenger from the queen watch the little man as he does stand sang and left, how he would soon have the daughter he always wanted because they would never guess that he's been mmos --
MOLLY: Want to lump will.
MILES: Will still scan, right. And the next day, the command came and passed now wishes quaking blockage by name. She said at first, "Uh ... are you called track." And no, Leon said. "But are you called a battery." Go and we had too. And what she said, "Then perhaps you and me is ..."
MOLLY: Rumbles caring.
MILES: And of the sound of his own name d old man was so furious studies stamped his fortune came the entire kingdom any broken to disappeared. The king McQueen, her daughter, lived happily ever after. (Pause.) Could make a good one. (Kisses Molly's forehead.) Street drains.
MOLLY: Yeah.
KEIKO: Bombing extension. (Kisses Molly's forehead.) The tax return.
(Miles and Keiko leave.)
MILES: To attack you into the bedtime story too.
KEIKO: Warm Olympic again Whitney.
MILES: Lecturing.

Pretty sure this is "If Wishes Were Horses".

SULU: Results Mozilla does investing.
SCOTTY: Transcriber result. But this jury wait awhile as long he's got to come on down camp.
SULU: Fifteen hundred miles in closing.
SPOCK: Transporter operational but just.
KIRK: Better be able to listen.

"The Doomsday Machine"?

O'BRIEN: K go on these past few days its final destination another one living in those waters. I want to set up a little workshop in the bedroom --
BASHIR: You set up a workshop in the bedroom.
O'BRIEN: Used much specific gruesome she's on trying to live like a bachelor again. Expressing subconscious desire push for a out about Walters outback.
BASHIR: These ridiculous.
O'BRIEN: So I said.
BASHIR: In anything by spending your free time in the bedroom accuplacer intimately associated with kanka you up actually expressing he decided not to be closer to her with during that since unit. Stock option really.
O'BRIEN: Exactly exactly. The warmest and quite catchy CDA white catching the more likely --
BASHIR: And --
O'BRIEN: The welcome mat.
BASHIR: Stone review and wish got was a map.

"Hippocratic Oath"?

 

[Trekker4747]

Hmmmmm... I wonder how YTTA would parse DS9's "Babel."

 

[Spot's Meow]

Can you name these Trek episodes as parsed by YouTube?

GEORDI: Had another idea that might help us to talk and personable Scandinavian.
DATA: Would be happy to.
GEORDI: Done some you're trying to teach the cat to use the consul.
DATA: The discharge for training program. And teaching sparked a jump down from the desk on my comment.
GEORDI: Has its own.
DATA: It's not entirely successful. Feelin supplement number two twenty one.

Force of Nature. Spot needs more screen time!

 

[tafkats]

Hmmmmm... I wonder how YTTA would parse DS9's "Babel."

Interestingly, it seems to yield different results for different uploads of the same scene.

KIRA: I suppose this is a good time to tell you the number three travel alert is broken down again. Junction.
O'BRIEN: Major marshall gardens.
KIRA: The.
O'BRIEN: That first go for the Muslim. Crucial to the secrecy.
KIRA: In not making any sense --
O'BRIEN: -- of humor and observe the quick. Close the reverse Harvard. Resected. Definitively.

KIRA: I suppose this is a good time to tell you the number three travel alert is broken down again. We just don't you.
O'BRIEN: Major marching band. But boards go for the new spending. Has shown to be seen plenty.
KIRA: You're not making any sense.
O'BRIEN: Remnants are going to bring. Crossed the river stalkers. Resected. Contingent upon.

Some lines are skipped entirely, including "Ankle try sound" in both versions.

"Reset gleaming" becomes "resected" both times, but "Dinner to bug" is "Definitively" in one and "Contingent upon" in another.

 

[scotpens]

I think Spambots use this feature before they post.

Asian translators who write English signage have been using it for years.