[Sigia-l] Re: Faceted Classification

[Christopher Fahey [askrom]]

> In ERIC, if memory serves me, wrong spellings in record entries were
> quite common and apparently even "accepted." One could see all the
> different permutations and go with the most popular spelling when
> making an entry. I suppose this "feature" was useful in that
> "synonyms" (misspellings) were collected that usually pointed pretty
> clearly to the descriptor.

"World Trade Center" and "Twin Towers" were, until recently, equally
popular terms for the buildings that used to define the NYC skyline.
Neither is a misspelling of the other, and the more popular of the two
may not have been the desired term. Without a controlled vocabulary,
there would be these two facets with the same content.

A better approach would be to "correct" each user's spelling by
cross-referencing each user entry with a controlled vocabulary. AFAIK,
the traditional method of ensuring adherance to a controlled vocabulary
in data entry is through a multiple choice interface element like a
pulldown menu. But your suggestions alludes to a better approach for
occasions when there are hundreds or even thousands of multiple choices:
if the user types "Twin Towers", prompt them with "Did you mean 'World
Trade Center'?

The "correction" process you describe may be more useful for the
searcher than it is for the data entry person. I think that this is what
they do at stock photo search sites: If you type in "businessperson",
the system can, for example, translate this term into controlled
vocabulary terms "businessman" and "businesswoman" and return
appropriate results. I think of this as a kind of pre-search
"substituter" to translate the wide variety of search inputs into the
small range of controlled vocabulary in the database itself.

-Cf

[christopher eli fahey]
art: http://www.graphpaper.com
sci: http://www.askrom.com
biz: http://www.behaviordesign.com