James Cridland's blog

A radio futurologist writing about what happens when radio and new platforms collide
This is an archive post from my old website. Not all links will work. For new posts, visit my main writing index.

« | Blog index | »

On-demand speech radio: disaggregation and transcripts

Posted on Saturday, April 9th, 2011 at 10:32 pm. #

The newsroom at DR

A while back, I went to chat with NPR. I was pretty impressed at what they were doing – particularly, their automated transcripts.

You see, the problem with a three-hour long piece of speech programming is that it is a fairly unsatisfactory experience when you want to listen-again. Chances are that you don’t want the whole of Morning Edition or the Today programme; you just want to hear “that report the man did about the dog”. So, when I go and see broadcasters across the world (like Belgium’s VRT last week or the Netherlands’ NPO earlier this year) I almost always cover the idea of disaggregation and transcripts.

Disaggregation splits a programme up from being an unweildy three-hour block of content into lots of discrete bits. This is hard work, and normally requires a human to do it (though there are ways of automating it); but well worthwhile, since it allows your listeners to find pieces of content that they’re interested in.

Transcripts sounds simple: just turning the audio into text. This means that you offer a choice of audio or text – different things please different people – but also means that you can make it easy for Google to index your content. More traffic, mostly from people who are unaware of your brand. Always a good thing.

I’m very excited to see a test page from Henrik Moltke which takes this idea one step further, using the audio from DR’s Hard Drive radio programme. (You’ll need a decent HTML5 browser, like Firefox v4, or Chrome).

This offers disaggregated audio and a transcript. It offers translation into English from the slightly less accessible Danish. Using HTML5, it also shows you the translation alongside the original language as it plays. Highlight a part of the transcript, and you can tweet that part automatically.

It’s a very impressive and strong example of what’s possible with new technologies.

Is this the future of radio? No.

Is it the future of taking content you’ve broadcast on radio, and making the most of it in an on-demand context? Looks like it to me.

Later: Here’s how it was built; this is a Mozilla project using popcorn.js.


Tuija Aalto
commenting at April 10th, 2011 at 6:50 am

Thanks for linking to this great demo! From the viewpoint of searchability alone this is worth exploring. Even in Finnish the speech-to-text is already decent enought to provide an index of key terms, if not a human readable transcript.

Henrik Moltke
commenting at June 2nd, 2011 at 11:57 am

Anyone wishing to explore this further, please get in touch with me (@moltke) or Mark Boas (@maboa)
Also, check out our Radiolab Demo, a collaboration between WNYC, Mozilla and Soundcloud: http://hyper-audio.org/r/


This is an archive site, and comments are now closed.