Posted by Libby Miller on

The highlight of this sprint for me was the multimodal/voice hack run by Nicky, Henry and Ant, with participants from IRFS and the Voice+AI teams from London and Salford. The idea was to experiment with extensions to voice interactions using different devices and ways to communicate (e.g. text messages, voice calls). Four teams each made an interaction prototype on the theme of dating.

I was particularly impressed with Ant's work on creating a radiodan-neue app with voice capabilities, using Rasa. Using that we were able to make a quick voice-operated Raspberry Pi radio, and some of the outputs of the hack will be demoable on a Pi.

Talking with Machines also ran a successful story development session (including a great role-playing exercise led by Joanna) with the two writers of The Next Episode which fed into stimulus for the hack.

Also in the Experiences team, Alicia and I are continuing to get the Tellybox prototypes ready for Taster, including a rethink of the interaction model for mobile.

I did a presentation about Better Radio Experiences at Supra Systems Studio's book launch (picture below by Tim).

Tristan has been wrapping up the NewNews project. He’s been trying to get some analytics on the pilots, writing a blog post and a video script. He also gave a talk about how to make new story formats at the Association of Online Publishers.

For standards, Chris is preparing for W3C's TPAC next week, and also hosted a Media and Entertainment Interest Group call on WebRTC, to discuss low-latency streaming.

The Disco team have been busy setting their direction for the next few months, after finishing up their work on the 'Made by Machine' programme.

They have decided to think about discovery more generally than just 'recommendations', 'personalisation' and 'search'. They will take a closer look at the existing ways in which our audiences discover our content, the role that our content plays in their lives, and to develop new means of supporting these needs and behaviours with our technology, focusing specifically on younger audiences.

Taking inspiration from Better Radio Experiences and NewNews, they plan to produce lots of diverse, tangible prototypes, developed alongside, and evaluated with, real people.

David and Jakub have been working on a pilot project - a 'BBC Mixtape' prototype which allows creative, thoughtful sharing of radio programmes, which will help us understand and prepare for the types of challenge we will face in later sprints.

In the Data team, Misa, Ben and Matt finished expanding the lexicon for the speech to text system. They have automatically added thousands of new words and their pronunciations, by analyzing BBC News and TV output and mapping various pronunciation dictionaries to the format the system requires.

This should allow the speech-to-text system to identify more recent words that have surfaced in the English language, such as 'Brexit' or 'Skripal'. They are now integrating this new lexicon into the speech to text system and testing its impact.

And finally - we said (a hopefully temporary) goodbye to Kristine, who’s gone to join the Data Insights team in Marketing and Audiences for a three month mini-placement, working on exploring applying some of her music recommendations work in BBC Sounds.


Netflix recently published about their work on HDR static images

There's an exhibition of computer art at the V&A

"I made a self-driving potato" 

This post is part of the Internet Research and Future Services section