In Internet and Society, David has been interviewing Sounds mix curators and Â鶹ԼÅÄ Introducing staff about how they might use the tool that we made for exploring the vast Introducing musical archive using automated tagging and similarity matching. He's had some fantastic feedback around which features work well and which don't. Also nice comments like:
"A production tool that has a sense of magic and unexpectedness about it"
Tristan, Henry and David have been running more workshops around better stock photos for machine learning /AI, and had a very interesting conversation with journalists about what their needs are when choosing a picture to illustrate a story. They've been planning the next steps for the project.
We published Ben's excellent blog post on "A machine's guide to bird watching".
Galen has been generating lots of ideas for prototypes and videos to explain ML & AI to 11-15 year olds, based on workshops he and Libby ran with the team. We've been looking at user research for the target age group (12-15 year olds), and with help from the team Galen's started to build two prototypes so far - see the screenshot of AI Art School below.
Background image designed by on | Easel Vectors by
Alicia has been writing her chatbot blogpost and creating an infographic to go with it, and learning about inclusive design and (even) more about SVG.
Henry's been improving our futuring tools and contacts and we're all still looking for signals and writing them up.
Libby and Galen enjoyed joining the data team's GPT3 hackweek.
Chris chaired his first meeting of the W3C Media Working Group, and joined W3C meetings on and WebID and federated identity.
We've all been planning ahead for the next 6 months.
The Interaction and Prototyping team have been working on version 2 of their collaborative game prototype - Mathieu has been battling some bugs, but there aren’t any major blockers. The team is planning to recruit external testers this month, both to test the prototype and also conduct individual interviews to explore our research questions. The team has also worked with colleagues (Lianne, Suzanne, Todd, Maxine) from the North lab to explore the best methodology for the testing in order to validate our hypothesis. The metrics developed from the Human Values research are going to be extremely useful as part of the testing methodology.
In parallel, the cross-Â鶹ԼÅÄ in-car strategy group is also ramping up work. Anthony has started exploring the technical opportunities as he’s now the R&D engineering representative in the Innovation Working Group, which Barbara is leading. She has started planning in collaboration with UX&D the structure for the first round of workshops/hackdays. The aim is to produce a series of demos for different in-car scenarios that can demonstrate the opportunities for the Â鶹ԼÅÄ in this domain.
In the Data team, the Speech-to-text team have been busy writing up their first draft of their X-vectors paper, prepping the repo to be open sourced and working with our legal teams around publication and open sourcing issues (neither of which are ever quite as straightforward as they seem). Matt has been diagnosing some slow run time problems with our enterprise system, Ollie has been digging into diarisation segment length and its effect on accuracy.
The Natural Language Processing team have been planning out new work on text summarisation for the spoken word/speech-to-text output and entertainment programme tagging, which they’ve just started. They also deployed the new Factual tagger to our colleagues in the Content Enrichment Chassis for trials. Ollie has been trying to get the speech summarisation pipeline working, Alex has been continuing his evaluation of , and Tamsin has mostly been tied up with interviewing.
The Recommendations team are currently involved in better understanding iPlayer user habits. Users visit at different times of day and at different frequencies, they also have different sizes of user history (i.e. have watched different amounts of Â鶹ԼÅÄ content on iPlayer) all of which can affect the ‘freshness’ of any recommendations we give them. The team are analysing this data as part of a six week deep-dive into understanding all these variables.
Ben has been helping out Joanna with her sentiment analysis and thinking about upcoming work with iRex. Joanna has been working with her PhD contact in Lisbon on the details of an app for automated sentiment capture and she’s working out how to do the next stage of her analysis of the sentiment data and finishing off the ‘Biometrics in the wild’ report. Polina has been carrying on her work on tagging, Kristine has been finishing off her Introducing work and Fearn has been creating a bunch of STM files to help extract X-vectors as part of her graduate project.
In the Data Science Research Partnership (DSRP), Andy has been busy working on the promotion and engagement of the partnership, continuing his work in raising the profile of our work both internally and amongst our academic partners. Alongside the monthly briefings he arranges he’s also been working on a workshop for DSRP members in July which should allow much deeper engagement with academics.
Links
.
- An Open Source Game Engine for Turn-Based Games