We are all used to matching apps that can help us find out which songs are playing on the radio, but researchers are also working on a similar technology that can help broadcasters and media companies to track metadata and find partial overlaps and duplicates — it’s known as partial matching. Patrick Aichroth, Group Manager at Fraunhofer IDMT, explains what it’s about.
What does the term “partial matching” mean?
With conventional matching, the goal is typically to identify content, such as a song or a video, using a snippet or sample of that material. With partial matching, on the other hand, your aim is to find any partial overlaps within a dataset, without knowing beforehand whether or where such overlaps exist. Both cases of matching may sound similar, but there are somewhat different technologies involved.
What can this technology do that matching cannot?
While matching enables the analysis of when and how often a piece of content has been broadcast, partial matching can analyze entire TV or radio programs or datasets and identify partial overlaps. This can be used for deduplication, to automatically propagate metadata within archives, or find out which parts of a preproduced item were actually broadcast, which is useful to create program cue sheets and clear rights. We have already carried out a few test runs with broadcasters and the results have been encouraging.
What does the future hold for partial matching?
We are currently running further tests and will then put the technology into a first product that can be used to detect and localize differences between preproduced material and broadcasted productions, which should be available by early 2021. Afterwards, we plan to provide further product versions for deduplication and metadata / rights tracking.