Thinkcorps

I watched “The King of Masks” this weekend, and I enjoyed it, even if it was a bit sappy. What was interesting was how overwhelmingly positive the online reviews were. When I thought about it, it became pretty clear that the only audience a film like this could ever attract is one that was explicitly looking for it. Thus, the positive reviews. As long as you weren’t disappointed, you’d be prone to leave a favorable review.

Likewise, negative reviews are usually left by people who have been misdirected into seeing a film, and thus found themselves in the wrong theater. They are the wrong audience, and in reading negative reviews it’s very difficult to see past the initial reaction that “obviously, this person was never meant to see this movie”.

So, my question is: what reviews are worth reading? Which should be weighted high, and which less? Because we’re working with a five star rating system, it seems critical to weed out the useless ratings which are skewing the results. But what algorithm could we use?

Collaborative filtering is going to match up a group of similar moviegoers and weight the reviewers within that group. The problem is, have you enough moviegoing history to get a good filter? If you do, is your group still large enough to give an effective answer? Usually not.

Anyway, no answers—but go check out audioscrobbler. If we could tap into that, link it to amazon and whatnot, we’d at least solve the music problem.