The University of Toronto is working to make video search easier via a new research project that will allow users to find related content and avoid copyright infringement.
Computer engineering graduate student Alex Karpenko has been working on Tiny Videos, a system that can compress video data and then search its contents. While the project is in its very early stages, he will be presenting his work at the International Symposium on Multimedia in California next week.
“A lot of people have experienced this problem with trying to search videos,” he said. “YouTube has over 100 million videos, and people often upload the same videos.”
This can make video search frustrating, as there are multiple copies of the same clips floating around, making it difficult to find different, or longer, clips when there’s no easy way to search for exactly what you want.
Such an improved search system could also provide better related search results. The current YouTube system, for instance, said Karpenko, doesn’t have the sharpest algorithm, often resulting in sub-par suggestions that are also filled with redundant videos. This would allow video producers to better index their videos.
David Fraser, an Internet and technology lawyer with the Halifax-based McInnes Cooper, said, “One of the issues is the big pile of data, which is hard to capture as video, and speech is hard to classify, and hopefully this would help with that.”
The Tiny Videos project involved collating 50,000 YouTube clips into a test database so that the researchers could find a way to search a large collection of videos and weed out what videos are the same, as well as what similar videos violate copyright.
Possible future uses of the project would be to generate reports of possible copyright-infringing videos so that they can be examined or removed.
“The No. 1 issue for companies that host video is copyright,” said Fraser. “Google is being sued by Viacom for its YouTube activities. It really is the next big Internet application, but the business models related to it have changed.”
One snag in using the Tiny Videos system to weed out copyright-infringing videos is the fact that videos that aren’t actually infringing copyright could get yanked unnecessarily. “A problem with automating any filter is the possibility of a false positive,” said Fraser.
This could mean an extra set of best practices for IT managers. Fraser recommends setting time aside to “take a more nuanced look” at the flagged footage to make sure that it is indeed infringing copyright.
It’s also important to educate users about the possible copyright infringement they might be practicing by making videos with popular songs backing them, for instance, or sharing copyrighted files at work. “Most of the time, it’s because people don’t know any better. Instead of just shutting down the technology, you should educate them,” he said. “You don’t want an overreliance on technology.”
