Segmenting broadcast content with The News Slicer
How might we automatically tag and segment broadcast news stories using existing production data and new technologies?
While the BBC is a leader in broadcast news, there are still challenges in efficiently using much of this content across the BBC's other output channels, especially online.
This project illuminates the ways we can use information from the production process to make it easier for journalists to discover audio and video content for use in other settings.
What does the News Slicer do?
- Hoovers up video and audio content from around the BBC.
- Segments broadcasts using production data such as running orders.
- Auto-tags each segment using subtitles.
- Makes the content available in "slices", with transcripts and linked data tags for each segment.
Why is this useful?
Some of the uses of segmented content are already understood by audiences, such as chapterisation. However with more sophisticated tagging it will be possible to build more bespoke and personalised news experiences, at scale.
Consider a commute podcast made up of segments that are combined to suit your travel time. Or a news bulletin made up exclusively of your favourite topics. Perhaps you desire several stories in a row about a topic area you want to go deep on.
Faster, more efficient production
For journalists, the ability to search through existing broadcast content makes it easier to use and re-use content. It is also easier to keep an eye on representation, for example whether our content is featuring equal numbers of men and women.
We worked with partners in BBC R&D to explore ways of making segmented content discoverable. Journalists responded favourably to the mock-up.
We want to build on this work to highlight and share the "best bits" of the BBC's broadcast content, before exploring ways of combining segments in novel formats for a more personalised experience.
- We have a working prototype available for use by journalists and a back-end pipeline that can be deployed for other applications.