Close ✕
How a major European football league leveraged AI to find key moments within its content.
Download Case Study

The Customer

A European sports broadcasting major responsible for production and distribution of match content for a major football league in Europe.

Business Challenge

The client had recordings and match content for all games of the league since its 50 years of inception. This included match content, several hours of raw footage, pre and post-match coverage, interviews, press coverage, analysis etc. 

Clients wanted to unlock monetization opportunities within this sea of content which was only possible if this archive is richly indexed with relevant data points.

Manual tagging was not a viable option because:

  • It was time-consuming, 
  • Did not deliver a high quality of media relevant data
  • It cannot possibly be used to cover the volume of content available.
  • It was expensive to perform

Key Highlights

Built for football

Specialized football metadata taxonomy to identify players/officials, camera angles, emotions, brand logos and other custom entities.

70% relevant tags via AI

As AI matures, its contribution began to reduce manual tagging efforts.

Auto-highlights

Auto-generation of highlights and promotional videos and publish custom content straight to the clients’ online platform.

85-97 % precision.

A managed service approach until the solution reaches the desired accuracy.

Our Approach

As generic API-based methods do not solve the specific media needs of the football industry, we were required to come up with a solution which:

  • Could effectively generate high-quality football relevant metadata.
  •  Could extensively cover the entire volume of content for enriching the video archive.
  • Utilized a niche taxonomy specific to football and custom to client needs, unlike generic API based methods.
  • Improved continuously over time and supports specific editorial tasks

Our Solution

#1

Hypertagging Engines for enriching content

a specialized AI toolset for football match logging that can generate metadata for enriching the existing content archives.

#2

Active Learning AI loop

The solution allowed loggers to continue with methods while the AI was learning their actions. After crossing a certain threshold of training data, the AI started recommending actions to the logger.

#3

Tailor-made editorial workflows

With Active Learning, every input provided by the logger improved the model over time as it adapted to any new information. With sufficient training data, archivers can create tailor-made workflows that address the specific editorial needs.

Case Study: Learn how a major European football league leveraged AI to find editorial moments within its content.