= XTech Talk Notes 17/06/06 = = Chopping Up Radio = == collaboratively annotating radio programmes == 1. TITLE Hello I'm Tristan Ferne, I lead a small R&D team for BBC R&Mi. R&Mi develop interactive offerings, from digital radio to the internet, for BBC Radio - a public service radio broadcaster in the UK with 10 radio stations. I'm going to talk about our Annotatable Audio project - in which we are developing a wiki-like internet application for annotating radio programmes. One thing to point out is that this is all R&D work, we are working on making this public but I can't guarantee anything yet But first a bit about why we're doing this... 2. DESCRIBING PROGRAMMES It's all about finding things. Extremely important for anyone that their content can be found on the internet. For web pages it's relatively easy and well understood because the content is basically text. But when your base content is audio or radio programmes it gets harder because the content is "encoded" as sounds, speech and music, so unlike web pages you can't easily search the content directly So we have metadata - data about data - that describes our radio programmes and is used in electronic programme guides (epgs), printed tv guides, web sites and search engines... Sometimes this isn't particularly useful as in this case of a music show on Radio 1... 3. DESCRIBING PROGRAMMES BETTER And sometimes it's a lot better... You can see that we now have a detailed description, some schedule information, repeat information, a mention of a related programme, information about the contributors... 4. CHOPPING UP RADIO One thing we don't have at the moment is metadata about how our programmes could be split up E.g. news - into stories or interviews magazine programmes - into articles or features interview - into speakers or questions or themes music - into tracks (controversial - rights/recording) for easier navigation or smaller, more convenient chunks of audio than whole programmes 5. ACQUIRING METADATA Now you might have thought that as the BBC we would have amazingly detailed metadata about programmes - because we created them but unfortunately that isn't necessarily the case. It's for a number of reasons - because we're a big and complex organisation and because production teams are there to make programmes and because we're got various unconnected IT systems and I could go on... But one of the things we do at Radio & Music Interactive is to try to solve this kind of problem; joining things up and extracting data to add value to systems. 3 ways of acquiring metadata: 1. Production process. From people involved in the production process or from technology used in the production process. E.g. our hard-disk playout system gives a feed about what is currently being played. Quick plug - You may have seen Matt Biddulph's recent write-up of when he was working at the BBC how he fed the now playing information for BBC 6Music to a secret user on last.fm 2. Media analysis. We can process the audio with computers to generate data; from low-level features like frequency bands or volume level to higher-level features like words using speech-to-text software 3. User annotation. Ask our audience to help generate metadata for the programmes they have listened to. One advantage we have here is a dedicated community - people really, really love some of our radio networks and programmes. As anybody who knows any Radio 3 or Archers fans could tell you. Note that stages 2 and 3 probably happen after the original broadcast. I'm going to concentrate on stage 3... 6. ANNOTATABLE AUDIO So that was a bit of an introduction as to the state of our metadata and why we want to generate it. The aim of this project is to get our listeners to divide our programmes into segments and to annotate and tag each segment We developed the original prototype last Autumn (Hi Tom!) to demonstrate the concept internally. We are currently in the process of refining the user interface and re-coding it so it is suitable for live deployment on bbc.co.uk 7. PLAY To help you get more of a grasp on what I'm talking about here's a demo... DEMO instructions This is an interview with Steve Reich (the composer) from an arts programme on BBC Radio 4 See the playback screen with a player Flash widget at the top, note the segments in green [play, skip through segments, drag playhead] Below this we can see user annotations for each segment (i.e. the green bits). You can see the current one being highlighted in red This makes an interesting and enhanced experience on top of a normal radio programme. 8. EDIT But what if you want to get involved... Well just hit "edit this" and we move into edit mode. Lets create a new segment where Steve Reich answers the question about his favourite pieces [add segment, playback question, PAUSE] Now we can see zoom in on the segment - possibly inspired by google maps! - and adjust the start and end points [play with zoom and segment] And we can annotate the segment... ["favourite pieces" "Music for 18 musicians" "Different Trains"] And tag it... ["reich","minimalism", "musicfor18musicians"] The annotation is based on a wiki-like model of editing so there is always just one single canonical version of the annotated programme. Anyone coming along and editing or annotating it automatically creates a new version of the programme. There is, of course, a history page to enable reversion to previous versions. I'm not demoing that cos I've only got the client running here. We're also toying with the idea of being able to add timed comments. You can see the "add comment" box on the playback page. Also would like to say that this is the current state of the user interface and is unlikely to be the final version. 9. TECHNICAL BIT Just a few details on the technical side of things. It's based around a REST-like API which serves up XML on nice, hackable URLs HTTP GET and POST are used to read and write to the server The client is based around a Flash app which provides the audio playback and rendering and Javascript/HTML for the annotation parts of the page. The Flash acts as a single point of contact with the server and communicates appropriate data with the rest of the page using the Macromedia Flash/Javascript integration Kit When we're done I'm looking at the possibility of making the code open source and also opening up the API for people to build stuff on. 10. COMMUNITY we already have a loyal and fairly active online community who contribute to messageboards based around programmes and radio networks, this should provide a good starting point for building a community around the audio annotation. But a wiki is going to be a very new experience for our site. We are intending to launch it around a relatively low profile programme so we start off small and it will probably be a factual programme as we want to promote the annotation angle rather than the discussion angle, which at the moment is covered by our messageboards. Users must be signed in to annotate or comment so there will be a small barrier to use. But anybody can see the marked up programme and use it as nav What do people get out of this... 11. WHAT COULD WE DO WITH THIS? We will have chapterised, annotated programmes - this extends and enhances the programme and makes it easier to navigate around and to find what you want to listen to We could provide a search that searches within programmes, so if you're interested in Steve Reich you can search for any segments or programmes mentioning him generate chapterised podcasts - these are currently supported by the latest iTunes and iPods using AAC and we would also like to support chapterised MP3s as recently added to ID3 standard These podcasts could be a nightly build based on the current, or approved, version of the chapters or potentially personalised chapters Or users could build their own programmes based on custom searches for keywords or tags and then the matching segments could then be concatanated into a new programme 12. TRACKLISTINGS We're also thinking about potential uses as an internal production tool - now we've got a nice, easy-to-use interface for chopping things up. E.g. many of our websites publish tracklistings for a show. This information may come out of our hard-disk playout system but for many specialist music shows this isn't used - music being played off vinyl or live sessions or shows being pre-recorded as a whole chunk. So we could use the annotatable audio interface to fill in the gaps turning what has previously been just text or maybe a word document into nicely structured data ready for re-use... 13. NEW INTERFACES Taking this area further we've recently been looking at some future interfaces for radio, metadata and chopped up programmes. Here's a a tag-based representation of a news bulletin. Each story, as segmented and annotated by the Annotatable Audio application, is represented by a tag, scaled to fit the length of the segment, so you get a textual representation of a programme through time... 14. NEW INTERFACES And here's a progression of that idea of using photos from flickr - the "most interesting" one for that tag. I built these interfaces in "processing" but I'm not brave enough to demo it now. Come and see me later if you want to see them. 15. SYNDICATION? Also thinking around this area, one thing we would like to do is make it easy for people on the internet to comment on, write about or link to our programmes. We've had this in mind for a while and our PIPs project, which the BBC Radios 3 & 4 websites are built upon, demonstrates this - unique, persistent URLs for every episode of every programme. But how about an easy way to include a clip of a radio programme on your blog? [blog this button in demo] There's a "blog this" button in the interface which provides a small fragment of HTML code, using an iframe, that you can paste into your website to point at this segment of the programme. You then get something like this... [slide] which is a tiny flash application embedded in your page which takes its data and audio off the bbc site or this... 16. SYNDICATION? Odeo do something very similar. There are all sorts of potential problems - particularly editorial and policy issues to do with our content appearing off site - but it's an aspiration for our department. 17. END That's it. I've shown our Annotatable Audio application how it provides a wiki-like interface for annotating and chopping up radio programmes, a bit about how we built it and some of our thoughts on where we could take this. There's my email address and my blog - keep an eye out on it for updates on this work. Thanks for listening, any questions?