I think the key isn't the algorithm but the data. You can do a decent
people-who-have-X-also-have-Y algorithm in a single SQL statement.
(That's how I teach MySQL.) But you need lots and lots of high-quality
data.
Various factors mitigate against this happening in the library world.
I think the math of the thing just doesn't support doing a decent
algorithm within a single library or system. But sharing
person-to-checkouts data across libraries raises privacy risks.*
Sharing data checkout-by-checkout, with no link between checkout
events, seems less likely to raise concerns, but is a much weaker
source of correlations.
I think, frankly, if you got 25 libraries to share—which might require
divine intervention—you'd still have something inferior to most the
better commercial systems.
I'll be interested to see if purely text-based systems, like the
recently announced BookLamp, will amount to much. I'm pretty sure they
will not. "Appliances" can't recommend books, only humans—whether
individually or though statistics.
*If you know six of my books from LibraryThing, you can probably
figure out my user name. Books are so numerous and many so obscure
that small numbers of books can point the finger quickly. A large
scale anonymized-person-to-book data transfer would resemble the AOL
disaster, where the released anonymous-people-to-queries.
On 3/14/08, Redirect-as-Sender <nancy.cochran_at_earthlink.net> wrote:
> It is my opinion that Amazon's "ability to recommend" is an exceptionally
> good marketing tool. Obviously, the algoritms can be written. Those same
> algorithms might be helpful to librarians for different reasons.
>
>
>
> Nancy Cochran
>
>
> > [Original Message]
> > From: Redirect-as-Sender <o.stephens_at_IMPERIAL.AC.UK>
>
> > To: <NGC4LIB_at_LISTSERV.ND.EDU>
>
> > Date: 3/13/2008 12:13:46 PM
>
> > Subject: Re: [NGC4LIB] food for thought and discussion
> >
>
> > Nancy wrote:
> > >
> > > I like to think that the thing that distinguishes librarians from
> > > technology -- whether that technology resides in an appliance or on a
> > > server -- is the librarians ability to select and recommend.
> >
>
> > What about "the thing that distinguishes Amazon from an OPAC is Amazon's
> > ability to recommend"?
> > :)
> > Owen
>
--
Check out my library at http://www.librarything.com/profile/timspalding
Received on Sat Mar 15 2008 - 00:21:23 EDT