Posted by Tim Cowlishaw on , last updated
Public Service Internet
It's been all change in the discovery team over the last few months - Olivier has left us for a new job as the Head of ODI labs - he's sorely missed, but we're pretty confident our paths will cross again in the future! In the meantime, the now-very-slimline team of David, Chris and myself have been continuing our work on our new Public Service Internet project that Olivier originally wrote about a few months back.
The main challenge we've found in getting this work started, is that the scope of the project is potentially enormous. There are infinite ways in which our public purposes and our public service remit and charter obligations can and should influence the work we do online - the challenge is identifying which of these are the most important, as well as being a good match for our skills and interests.
It's important also to consider the tangible outcomes we want from the project - what are we hoping to achieve by the time it's finished? To us, our main ambition is to understand better how an appreciation of the public purposes can be incorporated into the way we design and build digital products. Our public service remit is frequently discussed as something which applies primarily to the content we produce; we believe that designing our digital products and services with direct consideration of this remit can produce more innovative, useful and enjoyable experiences for our audiences, and which the BBC is uniquely positioned to provide. More concretely then, we'd like to produce prototypes of future products and services which show this possibility, as well as providing a methodological blueprint for incorporating our public service remit into the design process.
How then, might our public service remit influence the digital product design process at the BBC? This is another massively open-ended question, and one we spent a fair bit of effort trying to answer. Specifically, we've dedicated some time over the last two months to talking to our colleagues all over the BBC about the challenges and opportunities they perceive for the BBC as a public service provider of digital products and services. Reflecting back on everything we've learned in that process - there's four broad themes of 'Public Service Internet' activity within the BBC which we think that we as a team are well-placed to contribute to, and we've now got a practical plan for a project or set of projects which we think will allow us to do this. Look out for a longer post with more details on this next week!
Elsewhere in the team
The data team have been continuing their work on face recognition and speech to text: Denise has been implementing the face recognition stage of the paper on combining face recognition results with transcripts to identify unknown persons in footage, while Ben has been improving the runtime of the face recognition software to increase the amount of video we are able to process. On the speech-to-text project, Chrissy has been continuing testing the CTC method of speech-to-text, combining the results of the system we have built using Tensorflow with the Kaldi decoding tools we currently use.
Libby's been working through the project documentation, summarising the conclusions of our design-fiction research prototypes. She's also been experimenting with running iPlayer on a Raspberry Pi, with a view to producing some prototype set-top-box interfaces incorporating our earlier findings which can be used and evaluated in the home.
Talking with Machines
Joanne and Andrew have finished their report on Voice User Interfaces for children, and have shared their findings with the Childrens' department and the voice innovation group. Meanwhile the interactive radio experience is proceeding well - audio editing is underway and Anthony's been using our story builder engine to map out the first two scenes. He's also managed to get it working on Google Home, so we should be able to release the prototype for both Home and Alexa when it's done! Joanne's also been working with an MSc Student from UCL who is working on understanding the gap between users expectations of VUI devices, and their actual capabilities.
Libby, Joanne and Andrew have also been writing up some case studies about how we approach the process of research, prototyping and transfer, which we hope will be useful to others in the BBC.
Finally, Libby's written up and open-sourced her libbybot telepresence device which she uses to communicate with us while working remotely from Bristol. This was very timely, as Tom Howe has left us on attachment for a year to go and work for the Taster team up in Glasgow. Libby gave him the parts to make his own telepresence lamp before he left, so hopefully soon a Tom-bot will arrive at our office to keep Libbybot (and the rest of us) company!
- Google Design's guidelines on Human-centred Machine Learning.
- A lovely interactive essay on the game-theoretic foundations of trust.
- The background behind the body-positive, feminist, pedestrian crossings in Hilversum that Libby and I were admiring when we were there for TVX.
- Deep learning on a $79 USB stick
This post is part of the Internet Research and Future Services section