MovieLens
Movielens is a great site by the University of Minnesota that was recommended to me by Markos. It works like this: You rate movies you’ve seen and it makes recommendations based on your profile. The more you rate, the better the recommendations are. From the little time I’ve spent rating movies there, its recommendations are great: it clusters people based on their ratings and as such it starts recommending ‘good’ movies almost immediately. Within the first 5 minutes of my registration it started recommending movies I’d seen and that I’ve liked. This also accelelerates the learning process as you can rate the movies you’ve already seen much faster than a chonological, genre-based or other listing might allow. In about 40 minutes I managed about 250 movies I’d seen and liked, several times faster than what’s possible with IMdB.
If anything, MovieLens represents what IMdB sorely needs: a good recommendation service.
There’s also Criticker which works similarly: http://criticker.com/
Thanks for the pointer Konstantine. I hadn’t heard of it. Have you used it? Is it good in its data clustering/recommendations?
I’m probably not the most suitable person to ask since I’m not really a movie buff (thus, I haven’t used Criticker much).
However, from my limited experience I’ve noticed a “movies I’ve also watched keep coming up as I’m applying ratings” trend. In your MovieLens review you mentioned you noticed good movies you’ve already watched coming up; I’m talking more of a mixed bag (i.e. I was given movies I’ve already watched, but I certainly didn’t like all of them).
Do have a look at their site though; their raking & matching algorithms (TCIs, etc.) are explained and you can easily realize that the math/logic they’re using makes sense.
Oh, and they allow you to easily export your rankings in a variety of formats (XML, HTML, text) which makes me like them even more. I do realize that exporting your movie rankings isn’t exactly critical and it’s probably of no use (quite the contrary to, say, your email messages) but it tells something about the site’s philosophy — something that I like.
Well, MovieLens sorts movies based on its prediction of your grade, so, in this sense I did get some movies I’d watched but didn’t like too. The point I was trying to convey was that contrary to a chronological or genre-based list, MovieLens allows you to build your profile faster by showing you movies you might have watched (and typically liked, given that people rarely watch movies that they know that they’ll despise a priori).
Regarding your last point — exportability of the profile/rankings — I find it immensely important and one of the reason I do not use (or make minimal use of) services like Facebook, sync.gr etc., but prefer open solutions such as FOAF, RDF etc.