"Relevance" has been used in a particular way in computer science, and
in another way in library science/scholarship--and another in
Information Retrieval, which is sort of from library science but in
computer science. Actually, library science scholarship has evolved
different definitions over time too.
So let's not pretend there is some god-given "right" definition in the
sky that only Bernhard, or only library scientists, know! Yes, we choose
what words mean---this is the nature of technical terms and terms of art
used in scholarly fields, they are consensually agreed upon technical
definitions! Sometimes with more consensus than others, sometimes with
differing definitions in different fields. That's just the way it is.
So, the point of the kind of "relevance ranking" that new interfaces
(new to library world; this is not particularly new in general) --the
point is to try and improve the actual relevance of the results to the
user, based on what can be determined from the query entered and any
other context available. To call this "relevance ranking" is not to
claim that it does this _perfectly_ for all users---or that that's even
possible. Those who call this "relevance ranking" of course routinely
talk about how some algorithms for doing this may work better than
others, and it's an interesting question of how to best measure whether
a particular algorithm is better than another in a particular situation,
or overall--of how to choose one algorithm over another. Google's coup
was coming up with an algorithm which, in the web domain, worked better
for more users in more circumstances. But "worked better", yes, we mean,
gave users more relevant results the way Bernhard means it.
So Bernhard, I don't understand, are you suggesting that this kind of
software, attempts to provide algorithms to increase the relevance of
the results---is in fact a waste of time, and does not in fact result in
any gains for many users much of the time? I think the evidence in the
world would be against that. Users encounter information retrieval
systems in many more places than just the library now, and based on that
experience they are finding that our traditional OPACs without relevancy
ranking give many of them much of the time much less relevant results
(yes, relevant to their individual information need) than other systems
they use.
Or are you just saying that we should use a term other than "relevancy
ranking" for that? Do you have one to suggest? To me the most important
thing is that we define our terms, not neccesarily what term is used.
Jonathan
Ross Singer wrote:
> On Dec 21, 2007 9:54 AM, Bernhard Eversberg <ev_at_biblio.tu-bs.de> wrote:
>
>> So you are suggesting the POV of Humpty-Dumpty:
>> "When I use a word,' Humpty Dumpty said, in a rather scornful tone,' it
>> means just what I choose it to mean, neither more nor less."
>>
>> ... when using the word "relevance" with regard to search result
>> ranking.
>>
>>
>
> While this is a funny image, no.
>
> There is a "user expectation" of search and how it presents results
> and that our systems will try to choose an algorithm to present the
> most appropriate results based on the terms and context fed to it.
>
> I think, given the paucity of information available to it, determining
> relevance based on how the terms match based on predefined and ranked
> fields is a contract that the user can generally accept. The more
> semantics available, the more accurate the results. AquaBrowser is at
> least making the attempt at enhancing the semantics by incorporating
> more data.
>
> Since that's what we're really talking about -- it's relevance ranking
> of the metadata about items, not relevance ranking the items
> themselves and I think we need to realize that reality in how we model
> and generate our metadata accordingly.
>
> -Ross.
>
>
--
Jonathan Rochkind
Digital Services Software Engineer
The Sheridan Libraries
Johns Hopkins University
410.516.8886
rochkind (at) jhu.edu
Received on Fri Dec 21 2007 - 10:52:49 EST