Instead of a single dimension, my idea is that internet content consists of two things: keywords and emotional impact on the reader. So you tag the content, giving it semantic value, and then click on various indicators showing how it affected you emotionally.
So now I can provide you with associated material that is relevant and can predicatively adapt to your moods, i.e. responses.
It's cool stuff, but basically it boils down to a souped-up version of PageRank-like concepts.
I have a screen shot of a UI prototype but have to catch a plane right now. Glad to put it up if anybody would like to see it.
I was having visions of trying to parse sentence breaks, and then having users shove the good sentences in the article to the top, and bad ones down. None of which made any sense to me...
There is a bit of arm-waving going on here, I admit. But it's just a blue-sky concept right now, so a little arm-waving can be forgiven (hopefullly).
Aside from that question, what you describe sounds a bit like the online service SpinSpotter (which I think is a good idea, but severely underused).