Luke Gaydon, business development, sports at Accedo, on how he thinks AI will enhance sports content creation.
As sports broadcasting evolves, AI can enable content providers to step up their game. This is especially true for OTT services, where the technology can be used to tailor content suggestions, and even the entire user experience, based on past preferences and consumption behaviours.
AI will also play a critical role in enabling clip-based content to be pulled out of the archives, and could also ensure automated and effortless social media integration.
Using clips comes with a number of advantages:
- Clips are not as data-hungry as full matches and are therefore easier to deliver.
- Clips don’t need the same level of content security as full matches, especially if they are from old games.
- Clips are a great way of remembering and celebrating key sport moments, such as a record reached or a tournament won.
However, presenting clips is a very different challenge to presenting full-length programmes, primarily because of the need to curate the content.
This entails searching the archive for the right moments, editing them accordingly, and getting them into your video publishing workflow.
To date, the main barrier to publishing archive-sourced clips at scale has been the investment required to get it content-ready, which is to say digitised in the appropriate formats and with searchable metadata.
AI and machine learning have the potential to facilitate the fast and cost-efficient creation of a content-ready archive. Given the alternative is a long-winded manual process, being able to do this automatically is a game-changer for sports video providers looking to maximise the potential of their old content.
AI has already made metadata management and searching archives much more efficient, enabling automatic metadata tagging based on the video content. It also already works relatively well for automatic captioning and subtitling. For sports, however, we still have a long way to go.
Our vision at Accedo includes the ability to deliver a fully automated, accurate, heuristic, cost-effective, productised platform that works across multiple types of sports. Having done significant research into this area, it seems the industry still has some developing to do before our vision can become a reality.
We have so far not been able to find any AI platform that can easily switch from one sport to another without further training of the algorithm.
It also appears there is work left to be done in terms of detection accuracy. In some cases, the AI could automatically identify players with more than 90% accuracy, but we are still lacking the technology to successfully detect specific sporting moments. For instance, a type of goal from a particular position on the field, such as a right-footed shot from outside the penalty area.
Automatically identifying such specific events is a more complex process than, for example, facial recognition. Therefore, we will likely have to continue to rely on a manual checking process in the short-term.
That aside, technology has been developing at a significant pace. AI has a tremendous potential to transform the sport broadcast process, so we can expect there to be increased industry efforts to address the remaining challenges. Once we get the technology right, the opportunities it presents will be endless.
Luke Gaydon is business development sports at Accedo