using tropo to transcribe audio content

[Norman Khine]

hi your project looks very interesting.

perhaps this maybe of interest : http://blog.tropo.com/tag/transcription/

you can basically send a voice message and then have this transcribed. i have not tested it.

norman

[Dan Schultz]

Hi Norman et al,

  That's definitely a cool service.  The folks at the Center for Civic Media have even used it to build tools to make it possible to interface with websites via cell phone (see: VOIP Drupal).

  For the purposes of hyperaudio, however I'd warn that the quality just isn't there yet for anything that needs accuracy.  It might catch some percent of what you say, but that percent is far lower than you need for a transcript!

  There are services out there like SpeakerText and 3Play which use a hybrid Machine Learning + humans (Mechanical Turk) to try to reap the best of both worlds and even then quality remains a potential issue.  Maybe in 10 years we'll get there...

  On a slightly different note, has anyone seen tools or services out there that can take an existing transcript and time sync it?  That is a much easier problem from what I have heard, but I haven't seen anything that does it yet.  Maybe we should go implement it :D

Best,

 - Dan

-- 
Dan Schultz
Sent with Sparrow

[Julien Dorra]

Hi Dan,

Youtube can do an autosync for you if you upload both a video and an (english) transcript. It works pretty well, but if I remember well it's not word accurate, just phrase accurate, for captions 

You could also try to use Premiere Pros CS5 or 5.5. You can give it a "script" as an helper to the speech recognition engine. Each word is timed, and you supposedly can export an XML file (never tried it, so I might be wrong here). A good point is that you can download free additional speech engines for many different languages on the Adobe web site.

  On a slightly different note, has anyone seen tools or services out there that can take an existing transcript and time sync it?  That is a much easier problem from what I have heard, but I haven't seen anything that does it yet.  Maybe we should go implement it :D

Julien 

[Mark Boas]

Thanks Norman for that resource, for posterity I posted it on the pinned thread :  'Resources for Transcribing (voice to text)' https://groups.google.com/d/topic/hyperaudio/QjeItaQ1NeM/discussion

Dan - the guys at ramp.com provide audio-to-text (ATA) alignment of existing transcripts. It maybe worth speaking to them (copying Kyle into the conversation in case he wants to add something here).

Julien, Youtube's service certainly looks interesting, I wonder how we woudl approach this for audio rather than video - but it's good to have a free service available. Also I think you're right - Adobe are fairly active in this area as are Apple with final cut pro. It would be interesting to put aside time to try out some of these tools sometime.

Cheers

Mark