I’m going through the Netflix movie rating interface right now rating films & tv shows to try to improve my recommendations. Sometimes after you rate something it asks you how often you watch films of a certain category. Presumably the category the thing you just rated belonged to.
Some of the categories make sense, and some are completely insane. I can’t figure out how they could possibly be useful in determining what I like. Some good examples …
Parks and Recreation - NBC TV Shows
I don’t give a flying crap what network the show is on. I can think of several other categories it belongs to that seem more relevant (sitcom, documentary approach sitcom, etc.)
28 Days Later - Viral Plague Movies
Yes, I love those viral plague movies. But bacterial and fungal plague movies suck. Of course, what’s really silly is that this is obviously a zombie movie, not a plague movie. The mechanism of zombiefication is really not all that relevant.
Evil Dead - Demon Movies
Not horror, not comedy horror, not indie horror. Demon movies. Yes, because horror movies with non-demon monsters are so different.
Lust, Caution - WWII Movies
Ok, yes, technically, this is a movie set during World War II. Of course, when people think of WWII movies they generally think of movies about American involvement in the war, or maybe at a stretch things that happened in Europe. This is a movie set in China, in Chinese, with Chinese actors. I suspect most people think of this as simply a foreign movie or a Chinese movie.
Religulous - Spiritual Documentaries
Uh, WTF?
The Blair Witch Project - Wilderness Survival Movies
Ok, yes, they’re in the wilderness. And yes, they’re trying to survive. But again, WTF?
It seems like in many cases, they’ve identified some incidental feature of the thing in question and decided it is somehow meaningful.
Anyway, I’d love to hear comments from people who understand recommendation systems better. Do these categories actually make sense?
Comments
Zbigniew Lukasiak, on 2012-01-17 05:12, said:
Did you look inside these categories? Maybe the names are left overs from human created taxonomy, but their content is automatically determined and completely different from what you’d expect?