With the key goal of proving the potential value of conducting real-time, AI-driven content moderation on live broadcast and UGC content at scale and speed, this IBC Accelerator, led by Al Jazeera and Co-Championed by AP, Reuters, RTÉ and ETC created a PoC exploiting AI to flag inappropriate content as it happens on news & social feeds. The key innovation in this Accelerator is in leveraging AI-driven metadata for video, to perform automated content moderation in a live scenario for editorial news teams, professional and freelance broadcast journalists.

Champions: Al Jazeera (Project Lead), Reuters, Associated Press, RTÉ, ETC (Entertainment Technology Centre, University of Southern California)

Participants Metaliquid, TV Conal, TVU Networks

With the increasing volumes and critical importance of live content in the news life cycle, the boom in UGC on social platforms as a result of smartphone content production capabilities and the exponential growth of content produced by news and sports organisations are overwhelming human capabilities to manage, moderate and exploit.

While significant work has been focussed on automating technical QC for video content. This project and its proof of concept leveraged current developments in post-production content moderation, editorial QC and metadata enrichment to drive similar capabilities and value for live streamed content.

Bringing together the traditionally disparate functions of live technical QC, digital rights management and content moderation for editorial and regulatory compliance in a single workflow, this Accelerator demonstrated an AI-enabled toolset for media organisations to protect the integrity of their content and brand, the interests of their audiences.Specifically focused on detecting and flagging violent content in this phase of the project, further work continues to extend the potential scope of the project in later phases.

Why AI is needed for Live Content Moderation?
Today, every person with a mobile phone is a camera person, their videos can end up on social media, on broadcast television as UGC, and news people out shooting the news, of course themselves create thousands of hours of video every year.

The problem the industry has, and the challenge that this project addressed, is how to moderate that that overwhelming volume of video, to use automation and develop tools that can support broadcasters and agencies in deciding when video moves from camera to consumer at warp speed, what can and can’t be shown on air.

Automated content moderation in live production workflows, enables broadcasters to comply more accurately and effectively with their editorial standards obligations internally and with regulatory compliance externally, and allows producers to focus on their content rather than its management.

For further information: accelerators@ibc.org