I will watch that with interest. I have played with Netflix dataset, although not as much as I'd like. You can do a lot of useful things just with it, but in re

I will watch that with interest. I have played with Netflix dataset, although not as much as I'd like. You can do a lot of useful things just with it, but in real world it is very useful to use different definition of the problem from the Netflix one. "Similar items" are useful because customers can understand what it means, besides having some abstract "we think you would rate this X". Also it makes a lot of sense to integrate the other data you have about your items (genre, director). It is actually pretty easy to do.