IRFS Weeknotes #106
IRFS grew in size recently when we were joined by members of the Snippets team. Snippets is an internal production tool at the BBC with a growing user base and lofty ambitions. Rob Cooper, the team's producer explains:
"Thanks to iPlayer and its equivalents, finding TV and Radio shows has never been easier. But quickly finding content within programmes is almost impossible. We think that this is one of the next big challenges for broadcasters to solve, so we¹ve been prototyping ways to make it easier for programme makers (and ultimately the public) to find bits within shows.
BBC Snippets is part of this effort and it takes a three-pronged approach to the problem. Firstly - can you use the spoken content of a show to navigate? Because the BBC subtitles everything it broadcasts we can use subtitle transcripts as a way for users to find keywords within TV shows. This technique works particularly well if users know exactly what they're looking for.
Our second line of attack is object recognition. Some really impressive work has been coming out of academia over the last few years and we're starting work with a number of universities to apply the techniques they've developed to the wealth of archive content we've got in the BBC.
Screenshot of the Snippets application showing searching within a programme for a specific word.
And finally, we believe that sometimes easy-to-use tools can help solve problems that computational techniques can't. For example, if you're trying to find a particular sequence or shot and you know roughly when it occurred in a programme, then what you really need are quick-loading tools to allow you to scan the show by eye. A lot of our effort is going into developing these sort of lightweight scanning tools."
We are already starting to see some really exciting possibilities for combining the Snippets work with some of our other projects. We think the possibilities are really exciting and I'm sure you'll hear much more about it later in the year on this blog.
Now for a round up of some of our other projects:
Our work on unlocking archive content by making more effective use of metadata continues. Theo has been finalising the wireframes for an application which will help us to understand how and why people tag audio content. Andrew has been busy turning these into code. After this week the app allows people to agree and disagree with tags, and he's deployed the (Rails 3) app to an internal staging server. We're looking to put this in front of the Global Minds panel in the next couple of weeks.
Yves meanwhile has been working on a submission about ABC-IP to IBC. He's also continued on the automated contributor identification work. This uses speech models to identify individual speakers across the audio archive. He's been scaling it up so we can run it efficiently across all of our data.
The team have also been thinking about the next phase of the project. Dom arranged a meeting with the World Service to discuss needs and requirements. It's sounding like a very interesting proposal given the huge wealth of content at our disposal.
We're continuing to build some prototypes using the nascent Web Audio API. We've been using backbone.js, require.js and CoffeeScript to structure our applications. This week I started to document the code in a literate style using Groc and Matt was finishing off our ring modulator, gunshot noise and tape-loop simulators. Pete's giving the prototypes some exciting visual treatment while Oliver is tracking down some choice bits of archive content to give the prototypes some content.
Pete has been finishing up the UI design for the Google Chrome extension that will be used as part of the FI-Content study. Dan has been building said Chrome extension while Andrew has contributed some of the CSS. As we are targeting a single browser with this project it's great to be able to take advantage of that and use some interesting CSS3 features.
Meanwhile, Barbara's been working hard to finish up a long report on the project.
There's been a lot of activity around scoping our work for the next year this week. We had a good workshop on one of our themes "Discovering Media". Libby and Michael have been busy meeting colleagues from around the BBC and scoping various projects while Tristan and a few of the rest of the team meet with our colleagues over in Radio and Music to see if they have any interesting problems for us and how best to continue getting some RadioDNS-based technologies ready for a wider audience. And late on Friday Andrew and I got our little freerange printer to print a message! Expect more soon.
Here's some interesting things we found on the internet this week:
- RDFaudio is a Chrome extension which annotates /programmes track lists from radio shows with links to Spotify.
- MediaGoblin is a decentralised media sharing application which you can run on your own server.
- Bruno Maag discusses his bold quest to create a font that lets you say anything in almost any language.
- Our colleagues in News discuss Data Journalism at the BBC in the Data Journalism Handbook.