Amazon’s Emmy a Huge Win for Semantic Discovery
Earlier this week the National Academy of Television Arts & Sciences announced they will be giving Amazon Instant Video a technology & engineering Emmy award for its personalized recommendation algorithms. This is nice honor for the e-commerce giant, but the implications for the video industry are actually much deeper. The basis for awarding the honor was likely the historical impact that Amazon’s early item-to-item collaborative filtering algorithms, but the company may be soon heading toward a fundamentally different approach.
Collaborative filtering (CF) analyzes consumer data to find statistical connections between items and use that as the basis for recommendations. So if you put a new pair of running shoes in your basket, Amazon’s CF engine may recommend you a pair of running socks, since lots of other shoppers who bought running shoes also bought socks (Not just any socks, fancy running socks to go with your fancy running shoes).
Makes sense. It even works pretty well for consumer items traditionally sold by the e-commerce giant. When Amazon first launched its CF recommendations it was a huge innovation that was credited with contributing substantially to the company’s revenue growth.
But Amazon is using the Emmy award to flaunt its latest Video Finder service, that seems to leave CF behind and embrace a new semantic approach to recommendation. Amazon embracing semantics, a fundamentally different approach to recommendations, for its video content is somewhat like Ben & Jerry giving up ice cream. How could this be?
Amazon astutely realized that video is different. TV and movies are not regular consumer items. They are entertainment that is consumed based on personal tastes and our particular mood at the moment. The types of content each of us enjoy is not based on what ‘other people have also watched’, rather it has to do with the plots, moods, style and pace. So content has to be described and discovered the same way we choose and experience it.
Take a look at some of the categories in Amazon’s Emmy winning Video Finder service:
It includes classifications that describe the mood, plot, style and pace of titles. These meaningful classifications are the basis for semantic discovery. The implementation so far is basic and doesn’t yet include the awesome personalization that semantics can make possible, but the adoption of a semantic approach by Amazon is a turning point for the TV industry and a bold statement about the limitations of CF based recommendation technology.
For more about the semantic approach to video discovery, get your free copy of Super Search Me!
Watch the video series on Vimeo