Monday November 13, 2006

UnSuggestions

Every recommendation tool I've ever seen is trying to put the patron in contact with things that he or she will like based on prior consumption. This morning I saw a post on  LibraryThing's blog that did the opposite.
http://www.librarything.com/unsuggester

You can read all about it at
 http://www.librarything.com/blog/2006/11/booksuggester-and-unsuggester.php

Put in a title and see what other people who own or have read a work likely will never read, buy, or consider. Several examples are provided too (we learn that people who read The Confessions of St. Augustine do not like to read Night Pleasures (though perhaps based on the title they should in order to better understand just what the young Augustine was in fact confessing to)). And likewise. The algorithm is explained and it is quite fun for a few minutes to put in titles and quickly conclude that yeah, I'd never read those other things in a million years and whomever is reading them just has markedly different interests and tastes. It seems trivial at first glance, but it's worth pointing out that Tim Spaulding, the LibraryThing guy, posted earlier on NGG4LIB that

"I plan to use it to calculate "diversity" metrics
for users, and later various "levels." Basically you take a list of
books, eg., twenty top "academic" books and use their associations as
the touchstones that order all other books. I'm hoping it can produce
something like OCLC "Audience level" stats.
"

Data's fun.


Tim

Nowt that's interesting. And I wonder what will come of it.

On a personal level, LibraryThing seems quite neat, but I still haven't found it at all useful for professional work. The stuff I buy with my various collection funds is never found there. If I were in the public library world or possibly buying for the humanities, then I think the relevance might be higher. For my own reading, I already have so many things to read that honestly, the last thing I need are new recommendations (well, I'll take recommendations, but likely won't act on them). What I believe I enjoy most about LibraryThing is the FRBR-ization; seeing all the different covers and editions collocated in one place.

I'm left with two conclusions when reading the blog and Tim Spaulding's post. 1) A nagging  sense that OCLC just missed the boat here. and 2) That I agree wholeheartedly with Tim Spaulding. Data is fun.