I considered just such a rating scale when I was working on the Gaeabooks project, which was specifically going to use reader recommendations with links to buy books through existing bookstores, rewarding reviewers who did a good job describing and reviewing books. I was looking at the Otakuworld system of relating anime and manga by themes at the time, which did a reasonably good job of finding new anime and manga based on titles I already liked. I never did get the scales worked out, though.
The other way to do it is heuristically, by which I mean the method Librarything (and presumably Amazon) uses. This has the reader input a batch of their favorite books and/or authors, and makes recommendations based on other readers and what they listed as favorites (or bought, in the case of Amazon). Actually, I found that after I went through my Amazon account and fine-tuned my recommendations data (removing books and other items I'd bought as gifts, for example, adding books I'd bought through other sources, and rating the items I'd bought through Amazon) my recommendations started to get pretty good.
Somewhere in one of the boxes in the basement sits my copy of
A Reader's Guide to Science Fiction, an extraordinarily helpful book by Baird Searles, Martin Last, Beth Meacham, and Michael Franklin, published back in 1979, which listed major SF authors at the time, gave a brief description of their works and styles, and,
most helpfully, provided advice along the lines of "if you like this author, try these other authors." I found most of my favorite authors that way. I really wish they'd update this book, but when I wrote to the authors about this a few years ago, they'd all rather moved on to other projects since then.
I've seen a few websites try the same general idea over the past few years, but none that seemed to last (that I found). I'd welcome any suggestions in this area.
Oh, and Steve, for you, I'd make an exception.
