As you pointed out, WorldCat does all sorts of tricky ranking. I
believe there's a dashboard that they use for tuning the ranking.
Library holdings count, term frequencies, availability, FRBR, and
locality are all facets of that ranking.
In OCLC Research we do practically nothing without some sort of ranking.
In our VIAF project, we gather name authority records from 20-some
national libraries and merge matching records into a single VIAF record.
We rank search results by the size of the records, figuring that the
larger a record is, the more attention the component records got from
the national libraries and that size can be used as an indirect measure
of popularity.
In WorldCat Identities, we create author records from WorldCat data.
Simple SRU searches are ranked by the total number of items held in
libraries for that author. There is also a fuzzy name searching service
for WorldCat Identities that uses a combination of holdings and
similarity to rank results.
We use WorldCat holdings information for ranking wherever we can. For
instance, our FAST subject headings database returns results ranked by
holdings.
We've never done any usability testing on these ranking algorithms as
they are simply clearly superior to no ranking at all.
Ralph
> -----Original Message-----
> From: Code for Libraries [mailto:[log in to unmask]] On Behalf
Of
> Till Kinstler
>
> ...
>
> So, if you implemented something beyond term statistics based ranking,
> speak up and show.
|