I wonder if a reputation system could be applied to, say, Amazon book review ratings. It would be a more complicated calculation, but what if a star rating could be weighted according to those reviewers with whom I seem to agree? That data could be gathered by what I bought in the past and who reviewed it, and could be further refined if I gave my own rating later. Then instead of every rating counting equally, ratings for new books I'm considering are biased towards the opinions of those like-minded reviewers.
Mokalus of Borg
PS - Yes, it is complicated.
PPS - And I'm not totally sure it's necessary.
4 comments:
NetFlix does this, actually.
That makes sense. And I have a history of not doing my research very thoroughly before posting messages like this.
Sorry, I posted that in a hurry. No implied criticism meant. I just mean it's not as unfeasible as you might think.
Of course, with Amazon there are a few other wrinkles, not least of which is the fact that their dealing with an enormous amount of meta data. At least NetFlix is limited to a handful of video types, but how you go about judging what people will think of a brand shaving cream based on their ratings of a stack of computer books and a belgian waffle iron, I'm sure I don't know.
No criticism was implied, as far as I was concerned. You're in the clear, thud. ;)
And you're right again about the difficulty of mixed categories on Amazon. One idea would be just to let it happen and recommend waffle irons based on shaving cream preferences just to see how it works out. The simplest way I can imagine to limit the weirdness would be to section recommendations by category. But it still might be odd.
On a side note, seems Amazon's recommendations do get pretty strange sometimes.
Post a Comment