There's been some great progress this week on all three of our project strands, and plenty of exciting plans for future work!
Experiences team
The Talking with Machines team have been continuing work on our voice app, In particular, Andrew's been completing the scenario mapping exercise, and updating some of our scenarios to use a British English voice rather than American English (as this wasn't an option that was available to us before). The team have also created a short video demo of the voice app to share with other people around the ±«Óãtv. Happily this has gone down very well and has generated a lot of interest from people in other areas of the organisation - we hope to be able to share it publicly at some point soon!
Talking with Machines have also been working together with the Atomised Media team, planning a joint sprint on Conversational UIs for news which we'll be doing next week. The Atomised Media team also had a great start to the week, when an was featured on the ±«Óãtv News homepage and got upwards of 40,000 views in one day. They also reviewed the progress of the trial along with the Newsbeat and Taster teams which went very well, and are looking into extending the trial period further, since it's proved to be popular. Barbara, Joanne, Tim and Tristan have also been putting together a plan for further user testing of the prototype, and Barbara's been preparing to give a talk about the trial to a group of journalists visiting from the EBU. Tim's also been on a training course learning about the ±«Óãtv Digital Analytics system in order to crunch some of the numbers around usage of the prototype.
The Tellybox team have continued to analyse the results from their display at the V&A, and Tim and Libby have been planning a technical spike investigating ways of allowing groups of people to choose an iPlayer programme to watch together. Chris has been continuing with our standards work - Writing up his notes from the W3C TPAC conference, and preparing a presentation for EBU DevCon.
Discovery team
The discovery team decided to reduce the size of their backlog this week - therefore they've mostly concentrated their effort on editorial algorithms into three areas: shaping the design research for Smart Streams, getting better visibility of the deployment and state of the Content Analysis Pipeline, and continuing to merge the content analysis pipeline and database with Juicer. This has given them more time to deal with unplanned tasks and events - having a call with Google about their work, and maintaining open source projects. In addition, Olivier's been preparing for a on Web and Virtual Reality to be held later this week, and Fionntán's been researching techniques to extract information about the main protagonists of stories from text.
Data team
Ben's been working on Shot Detection using the , testing against a dataset. Jana's been investigating face recognition, following along with the paper 'Face Detection without Bells and Whistles'. Meanwhile, Denise has been generating MPEG-DASH test streams to emulate working with the API in a near live context, and Chrissy and Matt have been further testing Kaldi with some new datasets - one for radio, TED talks, and the famous .
So, in summary, plenty going on this week, across all of our projects. Happily we also managed to find time for those who weren't around for the shonkbot workshop back in to build robots, with help and encouragement from Libby. Above, you can see a picture of Katie's robot in action.
Interesting links
- - Summoning echoes and vibrations of git commits and pull requests to unwind our daily stress.
- - a great read on how changing habits of TV consumption have affected how we use TV to tell stories.
- - Interesting take on the politics of Silicon Valley's fixation with the 'simulation hypothesis'.