Netflix Prize
I registered to compete for the Netflix Prize. I don’t have a plan yet but I’ve got some ideas.
Marc at Orielly Radar and Greg Linden point out that mashups are probably going to be a popular strategy. IMDB and Amazon are the obvious complimentary datasets, but what about others?
For instance, I wonder what NetflixPrize + AOLDataLeak can produce.
Update:Oreilly Radar points to a Google Tech Talk by Dan Frankowski on- get this- how to identify individuals from anonymized movie ratings (he mashes the ratings up with non-anonymous movie message board posts).
Seems like AI would be pretty essential in doing something like that.
daveg
4 Oct 06 at 6:53 pm
AI is a pretty broad term, and means different things to different people. I try to avoid using it since it begs the question of what is “Intelligence” - a philosophical debate that drags me away from twiddling with technical bits.
“Data Mining” is closer, but still somewhat ambiguous. Like Neruolinguistic Programming, it’s more of a set of techniques than a coherent subject you can grok in its entirety.
banksean
4 Oct 06 at 8:11 pm
I meant, try using “programming” to solve this problem. :)
Yes, AI is a broad term, and I know little of the theory involved. the specific flavor I had in mind was the stuff Amazon employs for their recommendations.
I need to buy a book or two. It’s complicated stuff that I think can be used on a smaller, more everyday scale than it seems.
lets get drunk
daveg
4 Oct 06 at 9:54 pm
Oh I see. Yes, actually I was considering using collaborative filtering since there are several open source CF toolkits available and it shold be pretty trivial to hook them up to this data.
I have another more detailed post brewing. Will post it soon.
and I’m all about getting drunk. When and where?
banksean
4 Oct 06 at 10:09 pm
A warning - IMDB views competing for a prize as commercial use of their data, requiring a minimum $15,000 license.
Pete Kirkham
24 Dec 08 at 2:31 pm