-> trs24:

So your Master's project, eh? Building the better mousetrap/mp3 stereo? You'll have to keep us all posted on that. That's exactly the kinda thing I was hoping was heppening.

Will do. Be warned that this is purely a research project, though. We're coming up with ideas, writing requirements, doing some design work, etc... But none of us is actually smart enough (or wealthy enough) to make this into an actual product.

-> Bitt:

The TiVo is a good example. Just because I like Blade Runner and Alien doesn't mean that I want to watch every half-assed made for the SciFi channel movie. But the TiVo assumes I do, because all it knows is that I like SF movies. If there was some way to say that Blade Runner and Alien are good, but Code Name Phoenix is not, that would be a tremendous improvement.

Well it sounds like our system would figure that out, too. TiVo's flaw seems to be that they just look at one attribute (genre) and not multiple attributes (genre, series name, etc.) Going back to the music analogy, the TiVo example you've provided just sounds like an analog to the Pearl Jam / Creed example, just substitute the series name for the Artist. So if our methodology were applied to TiVo, you'd be rating your Blade Runner and Alien movies with high marks, and Code Name Phoenix would get low marks. Over time, your chances of being suggested Code Name Phoenix (and, to a lesser extent, other movies with attributes similar to Code Name Phoenix) would be reduced.

We don't have all the data structures/algorithms charted out yet, but it's starting to look like a neural network to me, and that's scaring the piss out of me.