To create the solution, human tagging was applied to a large labelled data set representing every item within a cross-section of Channel 4’s content. A score between 1 and 5 was applied for prominence and positivity on-screen. This labelled set of what valuable contextual moments look like in the broadcast stream was then used to apply image recognition and NLP to create labelling metadata on all C4 programmes. By comparing the metadata label with the human labels, contextual moments can be clipped, subjected to human review, then offered for brand advertising.
A controlled lab experiment was created to prove that this approach has an impact – in fact, the award-winning research demonstrated a doubling in ad recall and other key metrics. On this basis, advertisers were invited to participate in a live trial, with four being selected. This saw 300 brand-safe contextual moments delivered, based on scored quality of opportunity and an automated placement programme.
The solution is now being evaluated for full-scale roll-out with its potential impact is estimated in eight figures. Channel 4 has demonstrated that AI can be adopted in live broadcast media and not just digital channels.
