Posted by Tristan Ferne on
A fortnightly record of what BBC R&D's Internet Research & Future Services team is up to.
Internet and Society
Team Internet & Society are exploring what new forms of value that the BBC could provide online, we started 2020 by exploring whether could make online tools that help people choose and decide.
We have split up our team this sprint to finish up our work on ways how to help individuals with decision making. Team EXPLAINER will be making an interactive explainer (or "Explorable Explanations") to show what we've learnt and prototyped so far. Team PAPER are refining and making more paper-based prototypes. Paper-prototype are usually quicker to build and test than digital, and often better for thinking through things.
Following on from this we are going to start to look at how small groups of people choose and decide, starting with some user research with housemates.
We are preparing our study on digital memories for external publication, while talking to people around the BBC.
Team Anansi are developing media protoypes and experiments using emerging technologies. Right now they're exploring audio augment reality (AR) and voice assistant's synthetic voices.
Our survey of attitudes towards synthetic voices is out in the wild, now we're thinking of extending it. Please go and take the survey (and listen to some of our synthesised voices).
"Looking For Nigel", our first audio AR prototype is working towards public release, though Thomas and Henry are fighting Unity. We were hoping to take this to the SXSW festival, but that's now been cancelled. "Not A Robot", our second audio AR prototype, continues to be developed by Matt, with Joanna and Jack on sound design this sprint.
Team Data are developing tools and algorithms for media - including speech-to-text systems, sentiment analysis and programme recommendation algorithms.
In speech-to-text work, we've continued to improve bbc-speech-to-text, our implementation of speech-to-text optimised for the kinds of material and media we need to analyse, and evaluated some alternatives. Tamsin has started working on how to manage (and keep up to date) the vocabulary used by these kind of systems.
In recommendations work, Ben and Alex have completed their Apache Spark framework for evaluating recommendations at scale and are in conversations with BBC teams about them using it.
Joanna and Denise have been measuring their heartrates while watching TV programmes, preparing for a trial that will compare users' moods to the automated mood & sentiment analysis system we've developed.
Chris is looking at whether we can develop tools to help moderators moderate online forums and has increased the performance of his prototype abuse detection system. Next he'll look at some alternative neural network classifiers.
Next week the team will in a hackweek developing things around programme similarity and tag curation.
This post is part of the Internet Research and Future Services section