Ratings Systems
Way back when Netflix released a huge dataset for their competition, I got to thinking about ratings systems like the one Netflix has for movies that you rent.
You can log into Netflix and begin browsing movies and rate movies that that you like or dislike. For instance, I gave “The ‘Burbs” four out of five stars. Based on your ratings, netflix can better recommend movies that you might like.
A similar ratings system exists on the Yahoo Music player. As songs are playing you can rate them 1-4 stars or click on the “no smoking” sign which indicates that the song sucks really bad and you never want to hear it again (of course, there are ways that Yahoo will queue it up for you again, as I’ve heard “drops of jupiter” several times even though it makes my ears bleed).
I like the idea of these ratings systems because I’ve found a lot of music/movies that I like based on the recommendations of Yahoo and Netflix. What I don’t like about these ratings systems is that they have too many options.
I nearly always end up rating songs at 3 out of 4 stars on Yahoo (if I rate it at all), and I think this is my thought process:
- I like this song, what should I rate it?
- Well, its not the BEST song ever, so its not a 4
- On the other hand, I like the song, and a 2 seems like a bad rating
- I’ll go with a 3
At least to me, theres no reason to have any other option besides “Thumbs Up” or “Thumbs Down”, either I like the movie/song or I don’t. Digg got this part right on their site, you either “digg” a story, or you don’t, simple as that. I would think that simplifying the ratings system would make the recommendation engine a little easier too, since they don’t have to account for different “levels” of how much you like something.
As an aside, when I first started using the Yahoo music engine, I found that there was some way to change the rating system from a 4 star to a 0 - 100 system. 0 - 100??!?! “Gosh, I don’t know, is this song a 33 or a 34?”. Give me a break.
Shane Colin said,
Wrote on January 19, 2007 @ 9:53 pm
I went to this MISRC seminar at the U a couple years ago on this topic. Collaborative filtering is very interesting stuff.
Coincidentally, one of the earliest recommendation engines I ever saw was created at the UofM by a group of students. It was called MovieLens and it still exists today in some form. You create an account. Rate 15 movies, and it will in turn give you further recommendations. Good thing if you’re too cheap for Netflix.
Rett said,
Wrote on January 20, 2007 @ 9:23 am
I’m with you. In itunes the scale is 1-5 stars, songs I like a lot get a 4 and songs I friggin love get a 5. Everything else is useless because the way I see it, if I don’t like a song I’ll just delete it.
Have you ever used last.fm? I recently started using it and have found it a great way to track the music I listen to and learn about similar artists. I haven’t used the radio station feature yet but I think it will play songs based on what they think you’d like. When I’m listening to my songs I can click “love it,” as far as I know there isn’t a “hate it.”
ben said,
Wrote on January 20, 2007 @ 9:59 am
Shane - theres another project called GroupLens at the U now, I think its related to the MovieLens but more broad.
Rett - I’ll have to check that out, see if the recommendations are better than yahoo’s
Josh said,
Wrote on January 29, 2007 @ 9:10 am
I suggest you implement this type of ratings system for your blog. That way you’ll know what the readers like and dislike. Then you can change your life to better suit our tastes. Your blog would be way better if you only only did what your readers liked to read about. You’d have to quit your job (boring) and move to Vegas (hotbed of interesting stories).
ben said,
Wrote on January 30, 2007 @ 8:44 am
Josh, thats a good idea, I should also put out a virtual “tip jar” so that my readers can fund interesting stories. I can pretty much guarantee that the stories will get more interesting depending on how much money gets raised.