[Sigia-l] crosswalks thesauri - does context matter?

[Alexander Johannesen]

On 12/16/05, Eric Scheid <eric.scheid at ironclad.net.au> wrote:
> When conducting a crosswalk from one thesauri to another, does
> context matter?

I'm going to state the obvious, but yes, context always matter. I
guess your example needs more context, though. :)

When it comes to mapping from one ontology to the next (by indirection
or by derivation) there are nagging little semantic differences you
need to keep straight. Often this is referred to the upper and lower
problem of ontology work, where there is an upper ontology (in your
example I assume LCSH) and a lower ontology (which are local / closed
ontologies that somehow refines the "meaning" of the upper ontology,
but could even be DDC.)

As with anything else, it depends on so many things. To what degree of
difference between the datasets are we talking about before you can't
use the same upper ontology is always a great debate with no certain
answers, but it's safe to say that lower ontologies are normally not
cross-walk compatible unless the users are *very* similar.

> For example: I have an arts journal publisher client, they are tagging
> their journal indexes with terms from their private (specialist)
> thesauri,

This would be your lower ontology.

> and also tagging the articles with terms from LCSH.

Our upper ontology.

> They want to do a cross walk
> from LCSH to DDC, and have the corresponding DDC terms
> automatically included when they export articles tagged with
> LCSH terms.

Do you have this mapping? It is a fairly hairy, huge and substancial
one, and also one that cannot be correct 20% of the time especially
when lingustically mapped. Maybe it is this mapping you're asking
about?

> If another journal publisher wants to do the same thing, and
> assuming they have an overlap in LCSH terms (eg. metal-work
> arts and glass-work arts both use furnaces), would it be sensible
> for the second publisher to just assume the cross walk results
> of the first publisher, or would their different
> contexts result in different cross walks?

I guess the answer is that it depends on their lower ontology mappings
to the upper one, first of all. If you don't think that's a problem
(or it isn't a problem you'd like to tackle, and heck, I would
sympathise with that :), then it depends on the mapping done from LCSH
to DDC. As I said, straight lingustic mapping of the two are doomed as
a precise instrument but might be good enough for the assumption of
your question as both would be woring with the same imprecise mapping,
but if you've gota special mapping between LCSH and DDC you need to
scrutinize those mappings. Basically, what is that mapping based on?

Hope this helps a little,


Alex
--
"Ultimately, all things are known because you want to believe you know."
                                                         - Frank Herbert
__ http://shelter.nu/ __________________________________________________