F.A. Roos writes:
> Actually search engines are using metadata, albeit not the DC metadata.
> Ranking is performed in some way by some search engines, so it seems, by
> determining the matching of the search term vector with the index term vector
> of the "document", thus yielding a measure of relevance.
>
> BTW, sorting is a special case of ranking.
>
Should we talk about "relevance ranking" at all in metadata discussions?
True relevance ranking is impossible because relevance is subjective.
How's the computer to know what the user has in mind?
The fact that we often see interesting documents on top of "relevance
ranked" lists does not mean we actually get all the relevant stuff out.
In the absence of consistency, we can never be sure how much highly
relevant stuff is absent from our result set.
We in this list are all aware of that. It is the effect on the end-user
that worries me. We should avoid designing displays in a way that suggests
to them that the ranking has anything to do with relevance, but encourage
them to think for themselves.
I'm not saying a ranking based on some form of statistics is no use, only
that one has to be aware of its shortcomings.
Bernhard Eversberg
Universitaetsbibliothek, Postf. 3329,
D-38023 Braunschweig, Germany
Tel. +49 531 391-5026 , -5011 , FAX -5836
e-mail [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|