At Intel Research in 2005, I ran a study called "Voting with your Feet" that examined the relationship between a person's place preferences and their place visit behaviors. We were exploring a simple question: could a location-based system make recommendations about places that you might like based on your travel history? We looked at features like travel distance, place visit frequency, whether you visited the place alone or with others...
Our results were mixed (see UbiComp2006 paper here); it turns out, place visit behaviors are a fairly noisy signal when it comes to expressing an implicit preference. Although we tracked thousands of place visits, we only had the resources to follow 16 participants. Now, current technology like Yelp Mobile, Foursquare, and Google Latitude/Maps Mobile should make studying this question much easier. Indeed, Google now has "Google Map Suggest", which suggests places nearby that you may like. Although they are ambiguous about what features, exactly, they are using to make these recommendations, it's an interesting development for automated real-world recommendation systems.
No comments:
Post a Comment