TEXT OF INTERVIEW
Bob Moon: Do you rent movies online? One of the key components to the Netflix service is its ability to predict what movies members will enjoy by how they rate the movies they’ve already seen. Significantly improving upon the accuracy of these predictions has been a hurdle that Netflix has struggled with over the last decade.
So three years ago Netflix launched a contest challenging the public to help them out, a $1 million prize to the person (or team) who could come up with a way to increase the accuracy of the movie predicting algorithm by at least 10 percent.
Today Netflix announced that a team that called itself “BellKor’s Pragmatic Chaos” won the prize. Talking with us today are Chris Volinsky and Bob Bell, two of the members of that team. Welcome Chris and Bob.
CHRIS VOLINSKY: Hello.
Bob Bell: Hello, thank you.
Moon: Chris, let me start with you. How does your algorithm, your formula, translate into more accurate movie predictions?
VOLINSKY: Well, we’ve used all the data that Netflix released as part of the competition in order to build what are called collaborative filtering models, which predict what you’ll like, roughly based on the movies that your friends like. So if there is a person out there who has similar tastes as you do in movies, then we look at the other movies that they’ve liked and assume that you will like similar movies to what they’ve liked.
Moon: So I read that the Sandra Bullock movie “Ms. Congeniality” was the most frequently rated movie among Netflix users. That movie was savaged by the critics. Did you figure out why it’s drawn the most reviews from people who rented it. Chris?
VOLINSKY: That was one of the true mysteries of the competition. It turns out it was a movie that was very polarizing, suprisingly. People either loved it or hated it. Lots of people wanted to weigh in with some opinion, positive or negative, on that one.
Moon: You joined forces with previous competitors to create this seven-member winning team. You were from the U.S., Austria, Canada, Israel. How did you come together, and did you meet face to face to do your collaborating?
BELL: Well, we came together because our team was originally in first place and won the first progress prize. And during the second year we decided that our progress had slowed down, and that we needed to get together with other people who might have some different ideas, a different perspective on everything to help us make sort of a jump forward. The funny thing is that Chris and I had never met any of those members until either last night or this morning.
Moon: Are your brainiacs going to be putting your minds together for the next big Netflix challenge here? I understand they want to improve their recommendations for people like me who don’t take the time to rate the movies they want. How do you predict my taste in movies without my input?
BELL: A lot of what we’ve learned from the original Netflix prize suggests that they can really learn just from what you’ve rented, just from what … maybe even what’s been in your queue, or what you’ve clicked on and looked to find out more about a movie. So it’ll be interesting to see what happens in Netflix Prize Two.
Moon: And you guys are going to rise to that challenge?
VOLINSKY: We actually get pretty excited about these kind of big data sets. I know it sounds kinda geeky, but we like this kind of analysis, so we’ll probably take a look at it and see how much time it’s going to take to work on this. But first of all, we want to get some sleep.
Moon: Chris Volinsky and Bob Bell, two members of the team that won the $1 million Netflix challenge. Thanks very much for joining us.
Volinsky: Thank you.
Bell: Thank you.
There’s a lot happening in the world. Through it all, Marketplace is here for you.
You rely on Marketplace to break down the world’s events and tell you how it affects you in a fact-based, approachable way. We rely on your financial support to keep making that possible.
Your donation today powers the independent journalism that you rely on. For just $5/month, you can help sustain Marketplace so we can keep reporting on the things that matter to you.