2020 IASA - FIAT/IFTA Joint Conference

With the rise in content consumption and shift in viewership behaviour, football leagues are leveraging AI to log match information at a granular level for quick search and discovery within their vast content archives.

In order to meet this demand for high volumes of content, archivers use a standard process of logging match information and some are even creating match editorials and reviews. However, this process is manual reliant and can be optimized using AI.

AI will not disrupt the current setup, but be a layer atop these processes to drive efficiencies in speed, scale, and costs. Currently, the manual bandwidth becomes bottleneck for churning match logs with quick turnaround times, while API-based methods do not solve the media-specific needs. Moreover, there are no frameworks in the current model to utilize all this manual effort for training AI models. This is where Hypertagging Engines come into the picture - a specialized AI toolset for football match logging that enables the current processes with training data generation in the background. These engines will pave the way for a future-proofed technology by alleviating the technical debt created by the current systems.

In this presentation, we will explore how we have utilized AI-enabled Hypertagging Engines to generate metadata for enriching the archives of our premier client Sportcast GmbH.

With Hypertagging Engines, loggers can continue to do their jobs while  AI continually learns and improves the model over time through Active Learning. After crossing certain threshold of training data, AI starts recommending actions to the logger. With sufficient training data, one can create tailor-made workflows that address the specific needs of the archiver. Eventually, the redundant time spent by the loggers on a single file could now be spent scaling up to tackle large archives and unlock their vast latent monetization potential.