BNR Smart Radio (Round 3)
Summary
BNR Smart Radio aims to provide listeners with a personalized playlist, filled with relevant and automatically generated audio fragments based on BNR’s radio shows and podcasts. Users can tell us their interests and we will match them with a database of automatically segmented audio files with a rich set of metadata, generated by a model based on the latest AI/ML techniques.
This creates a non-linear and personalized listening experience for spoken word audio, catering to a new audience of younger listeners, while at the same time better serving our current listeners who are increasingly more interested in shorter bits of information instead of longer radio shows.
The solution
Up till now, radio in The Netherlands is mostly linear product. Finding a certain item after an episode has aired is almost impossible. By segmenting radio shows, we will be able to open our archive of radio shows to the general public. It will be easy to find short items, stay up to date with the latest items matching your interests, and to share these items with your friends and colleagues. Analyzing and segmenting all of our audio by hand, is a monumental task. We believe this time should be spent on journalism and creating more high-quality content. This project aims to completely automate the analysis and segmentation of audio, by using the latest AI/ML techniques.