> From: Code for Libraries [mailto:[log in to unmask]] On
> Behalf Of Jonathan Rochkind
> Sent: 14 May, 2007 10:00
> To: [log in to unmask]
> Subject: Re: [CODE4LIB] OCLC is "us" (was Re: [CODE4LIB] more
> metadata from xISBN)
>
> My understanding, from a number of sources, including
> comments Thom Hickey (I think that was Thom? I actually
> missed his name in my notes) made at the FRBR Implementer's
> Group meeting at ALA Midwinter, is that the algorithm current
> OCLC products use for work-set grouping is not limited to
> that published algorithm, but ends up being quite a bit more
> sophisticated than that. Thom (I think!) also said that
> different OCLC products used slightly different algorithms at
> the moment (for instance, FictionFinder vs. xISBN). I asked
> at the Implementor's Group meeting in December if OCLC
> planned to share the more sophisticated algorithms and tweaks
> currently being used, and he said he wasn't sure.
>
> It is of course OCLC's right as an organization to decide
> what information to share and when.

It is not a matter of sharing or not. For a number of projects and
some of OCLC's services we do in fact deviate from the published
algorithm. The deviation is a result to improve upon the published
algorithm. Sometimes when the published algorithm is tweaked it
produces slightly better or worst results.

I don't believe that we have significantly improved upon the published
algorithm and we will continue to experiment in our research efforts
and services we deliver to members. We published the algorithm so
others could use it as a base for their FRBR activities and encourage
others to experiment with our algorithm and publish their results,
especially where it fails to produce adequate results.

We are aware of a number of areas where the algorithm does not produce
adequate results, but have not found any resolution to those problematic
areas. I'm sure that we will publish any significant changes to the
algorithm in an appropriate journal when they happen.


Andy.