Jack says it well, as does Bernhard.
Alphabetic order is reasonably predictable for most people whose native
language uses latin characters. If you're plowing through a favorite
author's books, you can check holdings against the list you brought with you
(or that's in your other browser window). Does that make it the "best" sort
order? Not necessarily--but "relevance" for bibliographic records is, I
believe, a chimera, and last-in-first-out is frequently a mystery.
Ideally, and in many systems realistically, a user should have a choice of
sorts, including chronological/reverse chronological--but the default should
be set with some sensitivity, since the default is what most users will see.
My own belief (and what we do/did in Eureka, after loads of user feedback,
looking at all the studies we could find, etc., etc.) is that, in general,
reverse-chronological (newest first) probably makes sense for articles and
alphabetic (by author, by title within author) probably makes sense for
"whole items" (books, sound recordings, movies). As defaults.
"Most popular" and "most relevant" are assuredly not the same thing, even
though that's part of how some engines operate. I prepared a thought
algorithm for crude "relevance clumping" in articles without full text last
year. Since it's certainly not going to get implemented in Eureka at this
point, I should look into whether it's considered proprietary information.
But even that clumping is, to some extent, silly.
walt crawford
On 6/14/06, Bernhard Eversberg <ev_at_biblio.tu-bs.de> wrote:
>
> Jack Hall schrieb:
>
> > I agree with Alex that relevance to users is a slippery slope, in terms
> of
> > how we can predict or ascertain it.
>
> Let's face it - we cannot. Relevance is tied to meaning, and no
> software has access to that, which would be to the user's actual
> thinking and intentions. From the two or three words someone puts
> into the slot, no software (and no human either) can reliably guess
> what they wanted.
>
> Relevance ranking is something alien, soberly looking at it, since
> it is strictly formal and algorithmically derived - software to this
> day can do nothing better, they are all sequential machines
> manipulating character strings. One may distinguish between the
> purely algorithmical and the dictionary based procedures, but the latter
> still don't make a "thinking machine" out of a computer.
>
> Corollary:
> A good catalog will avoid misleading vocabulary like "relevance"!
> At least, avoid the use of such terms without providing an explanation.
> Another example would be "dialog".
>
> Regards, B. Eversberg
>
Received on Wed Jun 14 2006 - 11:08:56 EDT