Posted by BBC Research and Development on , last updated
It’s been a big year in BBC News Labs writes Allison Shultes. We’ve delivered a number of products, from graphics reversioning tools to news bots, and we’ve worked with teams from the World Service to Visual Journalism to R&D to drive innovation forward within BBC News.
It’s been a big year in BBC News Labs. We’ve delivered a number of products, from graphics reversioning tools to news bots, and we’ve worked with teams from the World Service to Visual Journalism to R&D to drive innovation forward within BBC News.
A big focus of News Labs in 2016 has been video reversioning and production workflows. We’re closing the year with a finished ALTO web application – a video translation workflow that allows journalists to reversion videos using text-to-speech technologies and synthesised voices. We added new features that give the user more control over the pronunciation of the synthetic voices and the intelligibility of sentences. We’re getting ready to hand over the tool to production, where it will be maintained after being introduced into journalists’ day-to-day workflows.
We’ve also begun successfully introducing our GraFix tool to Language Service teams, allowing producers to easily repurpose graphics templates and social videos with text overlays directly from their browsers. This product speeds up the process of reversioning English-language graphics templates into a target language. Feedback from the head of the BBC Japanese news site, one of the first places where GraFix was introduced, has said that the tool will be “very, very useful” going forward, and we’ve heard from other World Service producers that GraFix has enabled even faster and more efficient workflows for BBC journalists.
Our GraFix tool lets producers easily reversion graphics templates from their browsers.
In the realm of video tools for production, we’ve made significant headway on our Online Content Toolkit (OCTO), a project exploring media editing, repurposing and augmentation based on the results of intelligent analysis. We're using enrichment processes like automatic speech transcription, concept extraction and shot detection on BBC News production content and aggregating the results in a central repository, which then forms the basis of specialised tools. Going forward, we’re interested in exploring the automated generation of subtitles and repurposing of video content for users on low-bandwidth connections.
Our bot work has also been a success for our team. In November we launched the BBC Mundo Messenger bot in collaboration with BBC Mundo and the World Service — the corporation’s first news bot for Facebook Messenger. We plan to continue implementing new analytics in order to get a better understanding of how our audiences interact with a news bot service.
The BBC Mundo Messenger bot launched this October.
We also partnered with Visual Journalism to make their graphics workflow on the night of major election votes more efficient. In the summer, we built a Twitter bot that automatically repurposes television graphics into shareable social media cards reflecting the most recent vote count in the EU Referendum. This autumn, we repurposed the original code so it could be used during the U.S. presidential election. We plan to hand off the product to Visual Journalism so that they can manage it without our support during elections in the future.
Twitter graphics generated by our EU Referendum-turned-U.S. Election bot.
We’re still keeping our eyes on conversational user interfaces as we head into the new year. We’ve also gained experience developing more ambitious prototypes at hackathons that have introduced us to new technologies. A team of Labbers won the IBM Watson Build a Bot hackathon with an experimental pitch using a combination of text-to-speech and personality analysis services. The 100 Women Hack held by Connected Studio allowed team members to experiment with multilingual API services and automatic graphics generation based on user input. We hope that these experiments might inform our work on bots in the future.
News Labbers Lei He, Alex Norton and Rachel Wilson at the IBM Watson Build a Bot Hackathon.
Even when trials of our projects haven’t gone as we expected this year, we’ve been able to learn from the process. Built in collaboration with BBC R&D, our automated External Links Manager (ELMer) attempts to generate a higher number of onward journeys from BBC News story pages than our current solution from a 3rd Party system. While a recent trial suggests that we still have more work to do, it also spurred our first use of multivariate testing on BBC News story pages and consequently gives us an understanding of how to implement this type of testing in the future.
News Labs has continued its support of hack events that bring together journalists and developers to solve problems around news. We started off the year with two University Challenge events at the London College of Communication and City University. We also held a smartphone storytelling workshop for the Radio 1 Academy in Exeter, following off of a successful event in 2015. This autumn, we partnered with Connected Studio and The Trust Project, which is run out of Santa Clara University and funded by Google, to put on the Trust Project Challenge — a two-day hack to prototype solutions for building trust in online news attended by 60 journalists from organisations including the Guardian, Buzzfeed, La Stampa and the BBC.
News Labbers host a University Challenge day at City University.
Our early plans for the new year include further exploring the potential for chatbot technologies for news, producing more tools for creating new types of content, exploring more atomised forms of storytelling and continuing to run hack days for students, journalists and developers.
More on News Labs: