Yesterday, AbeBooks and LibraryThing announced the rollout of "BookHints", a recommendation system based on LT user-library data. "When you now search for a book on AbeBooks and click through to a listing, you are presented with up to six recommendations for books you might also wish to read. BookHints generates recommendations based upon titles found on the bookshelves of like-minded readers who also own the book originally sought," summarizes Richard Davies at Reading Copy. Currently only about 10% of listings will have recommendations, but they're expanding rapidly, Davies notes.
Over at the LT Blog, Tim's got some more background on how this all works.
I think this will do much to improve AbeBooks' "browsability factor" - while I usually just go there if I'm looking for something specific, it's always useful to have good recommendations ... and from what I've seen so far, LT's various recommendation algorithms are about as good as they come these days. Yes, that means they're better than Amazon's. GalleyCat pooh-poohed BookHints, saying that the recommendations are "basically the same thing as Amazon.com's 'customers who bought this item also bought...' recommendations, only with book jackets! And based on the data from LibraryThing members rather than actual sales data, of course."
At least the author there notes a comeback from Max Magee of The Millions (link added to sidebar), who says "You shouldn't underestimate how much better LT's recs are than Amazon's... People buy books for many, many different reasons - as gifts, for work, etc. However, a book that is part of one's library is a totally different thing; it's most likely there because you enjoyed it or expect to." Exactly right. My recommendations on Amazon are always annoyingly skewed because of gifts I've bought (sitcom DVDs for my sister, children's books for my young cousins, &c.) - on LT, I don't have that problem. And neither will AbeBooks users.