I just came across a fascinating article about how Netflix categorizes movies. Did you know, for example, that Netflix actually pays over 40 people a couple hundred dollars a week (each) to watch 5 or so movies (per week) for them and tag each movie in great detail? How do I get a gig like that? :-)
Basically, the taggers watch the film and keep track of all sorts of details--happy or sad ending, amount of profanity, humor, quirky, sci-fi, romance--you name it. Each movie then has more than 100 different data points, and those are then used to recommend to Netflix viewers what they'll like and what they should see next.
I enjoyed the article on many different levels. First of all, as a Netflix user myself, it was fun to get an insight into how the company decides what I'll like and how much I'll like it. Yes, a fancy algorithm is the big basis of this, but you need to put data into that algorithm, and I'm sure this is some of the data that piece of tech needs to function. It's nice to know that all the fancy computer programs in the world still can't substitute for human beings watching something and evaluating it.
Second, the librarian in me loved seeing the care and detail a company is taking to make sure it categorizes its offerings correctly. There's really little difference between this and what I do as a cataloger. The article mentions how some movies can be harder to categorize than others, and that's true about any classification system. You'll have something in place that seems like it covers all the bases, but there's always going to be some strange entry that doesn't really fit anywhere. The solution is to force it somewhere--because if you keep making crazy exceptions for individual items, then pretty soon you lose the utility of the organizational system you're using.
Anyway--just interesting to see a "real world" application of some basic cataloging principles, used for a purpose that so many people take advantage of every day. Wanted to share. :-)
No comments:
Post a Comment