At VideoMind, we're always thinking about video—and how media companies and brands use it to entertain, engage and communicate with audiences. But we're taking a step back to look at how video affects brain development and cognition—what happens behind the eyeballs, in other words. This is the first installment of an occasional series. Read the first here.
The Internet is a sea of videos. With eight years of content uploaded to YouTube
each day, it's hard to find what's interesting or relevant to you.
As it stands, content recommendation engines are just beginning to live up to their full potential -- there’s plenty of room for improvement. To build more-robust engines, engineers can incorporate users' social graphs, socioeconomic data and a host of other data. BBC’s research team
is taking its engine in a new direction, testing out ways to incorporate users’ mood to improve its content discovery model.
The media company’s research and development team is exploring beyond rigid TV and movie genres. Instead of searching metadata, such as title or director, it’s looking at more subjective content discovery. Its research blog explains:
[I]f the BBC archive is ever to be made available to the public, we’re going to need some help finding what we want. From hundreds of thousands of hours of programmes spanning over 75 years, simply searching for ‘comedy’ isn’t going to get you very far! In fact, in our recent study, the majority of people said they would find it useful to be able to search by mood. I understand if you’re sceptical, we’re so used to conventional searching, it’s hard to imagine a useful alternative. However we’re not suggesting this approach will replace conventional methods, more that it will augment and improve them. It may even turn out that searching by mood is something people didn’t realise they wanted it, but once it’s there they’ll wonder how they ever managed without it!
For now, the BBC envisions this classification system to analyze video and audio features such as luminosity, laughter and motion. Using this information, readers can find content based on the progression of a movie's mood.
For example, a programme with a high level of motion but not much laughter might score 5/5 on the ‘slow-moving to fast-paced scale’, but 1/5 on the ‘serious to humorous’ scale, meaning it is quick but not very funny (a thrilling drama for instance).
With this infrastructure in place, algorithms can recommend programs with similar mood fingerprints to content they like. Creating a mood chart is still a work in progress, but the current prototype interface is a two-dimensional scatter chart, with each axis representing a mood scale. With this chart, users can choose a particular mood using sliders or by searching for terms, such as excitement.
Content discovery is the next big frontier for online video -- one that Ooyala's Vice President of Engineering Nimrod Hoofien mentioned as an area that desperately needs attention
. BBC attempts to find a unique and outside-the-box solution. Using the idea of mood discovery, the final product can take on a number of different shapes, and we’re “excited” (get it?) to see what comes from this project.