We’ve had web-translation for yonks. But given that podcasting is often an inefficient way to both take in and disseminate information I’d like to be able to go to a site and feed in a relevant audio or audio-visual file – or point the site to a YouTube video for instance and get a first draft of a transcript. (I know it would have some errors, but they’d be amusing in any event – nothing like throwing a bit of randomness into the mix).
I thought such a service ought to exist. Perhaps it does, in which case please let me and other Troppodillians know in comments. If not, I know you’re listening, Larry, Sergey, Steve, and Bill (well actually I know Bill’s not listening), so lets see it – quick smart.
And in case you’re interested, I wanted to get transcriptions of this video (HT Joshua Gans) and this podcast so I could blog about them. Both pretty interesting to me anyway.
PS: and why don’t sites like YouTube and TED allow me to download mp3 files of their videos, at least where they’re lectures, like blogging heads does?
You can get software to rip an mp3 file off streaming video but don’t ask me for a link cos I’m too lazy to find it.
I doubt that you’ll find any worthwhile transcription software. Voice recognition software like Dragon Naturally Speaking has to be ‘trained’ to recognise your voice and even then it has a high error rate. Using it for a dodgy sound quality lecture or similar would produce meaningless gibberish I suspect.
Take it from a linguist: the only good way to transcribe anything is to do it yourself. No such software exists apart from that which will play a video or sound file and allow you to seamlessly type the transcript and timecode it. Useful software though. It’s called Elan and is used by linguists everywhere.
Ken, Jangari,
You’re being too negative.
Training voice recognition software is exactly what could be done using Web 2.0. So I ask the software to transcribe the first 10 minutes of Sam Rosenberg – the first link I linked to – or Clay Shirky – the second one. I then train it by marking what it’s come up with. Bingo, I’ve then helped train the software in Sam Rosenberg’s voice. Perhaps Sam could get an email from the site, saying that it’s now trained in his voice. But whether or not he does, it would create a capability.
Then if someone else wants to get Sam’s voice transcribed – or Clay’s or anyone else’s – they search the transcription site to identify whether it’s learned that voice or not and if it has it does it for you, and if it hasn’t it offers to do it badly and give you an opportunity to train it.
And after that, something that will alter it enough to submit as an assignment without being caught, possibly also listing how many layers of hegemony it has.
Geez, Nicholas…..you’re not wanting much, are you.
When computers develop the ability to transcribe automatically, the world will change.
Aside from anything else, the security industry will be so happy its brain will explode.
Tape, transcribe, search, accuse …
I’ve just cleaned up an old manual typewriter. Now I am boggled about how much easier writing has already become. But if I was just disseminating cleaned up conversation…
CastingWords uses Amazon’s Mechanical Turk (ie, real people) to offer transcription for $0.75/minute, or more if you want it quickly.