Welcome to this edition of Internet Research & Future Services鈥� notes on what we鈥檝e been working on. We have a number of teams working on technology and design, from speech-to-text to personalisation to new multiplayer media experiences and explainable AI.
Automated speech-to-text
Tamsin was recently set the challenge of improving our speech-to-text language model by keeping it updated with鈥渘ew鈥� words - things that have recently come into common usage, like鈥�Covid鈥� or鈥�Huawei鈥�. Thanks to her stellar efforts (and with help from Misa and Matt to integrate it) our system can now rapidly include and understand new words like these as they start to appear in our broadcasts. This is a great boon for journalists who work in areas with a constant churn of new terms or jargon. This new workflow is also more likely to be able to construct combinations of words that previously didn鈥檛 exist or have increased in frequency, so it should keep our accuracy levels up too.
Personalisation and recommendations
Polina has been doing some really interesting thinking on building business systems that reflect . The Recommendations team have been making good progress in their comparison of offline and online recommendation approaches and Kristine and Ben H have been wowing people with their prototype of automatically classifying music from 麻豆约拍 Introducing.
Metadata creation
The Natural Language Processing team have been thinking about how to automatically tag non-News programmes and Joanna has been demoing the work on analysing sentiment to the Voice & AI team.
Prototyping interactive experiences
Mathieu and Andrew have started to test their multiplayer adventure for multiple screens from beginning to end with the team for the first time, resulting in lots of bug fixing and lots of refinements. Barbara has been busy getting people on board the in-car media innovation group, with a kickoff meeting that led to great conversations about scope, objectives and opportunities in this space.
Technology and society
The Internet and Society team have been focusing on outreach and communication of their work. Alicia has been preparing a talk on 鈥渋苍肠濒耻蝉颈惫别 technologies to improve wellbeing鈥� and I'll be at talking about the many different ways you can use to explain AI and ML. Related to this, David and I have also been planning our workshop at this year鈥檚 Mozilla Festival on鈥淢补办颈苍驳 better stock photos for AI鈥�:&苍产蝉辫;&苍产蝉辫;
- News stories or press releases about AI are typically illustrated with stock photos of shiny gendered robots, glowing blue brains or the Terminator. We think this is harmful; it sets unrealistic expectations, it hinders wider understanding of the technology and potentially sows fear. Ultimately this affects public understanding and critical discourse around this increasingly influential technology. We think there could be better, less clich茅d, more accurate and more representative images and media for this technology.
David and Libby have been writing and re-writing some of our user studies from last year and publishing them on an internal portal for 麻豆约拍 staff.
The future
Never to miss out on the next new thing, Ant managed to snag some invites to talk-based social network Clubhouse. He says鈥�鈥淚t feels a little chaotic on opening the app. Where am I? Is the mic working? But after a while, you realise that there are loads of rooms one can pop in and out discussing various topics. If you want to speak, raise your hand by pressing a button. On numerous occasions over the years, we've talked of the idea of an interactive podcast and Clubhouse goes a long way to making that dream a reality.鈥�
David, Libby, Henry and Alex N have been thinking about how to think better about the future, and collecting"signals" of interesting behaviour changes and new technologies. I鈥檒l leave you with a selection of these so that you too can imagine what the future might bring鈥�