[Sigia-l] Re: Facets vs. ontologies [was: Sigia-l] Findability - hierarchies

[Lars Marius Garshol]

* Lars Marius Garshol
|
| Still, the fact remains that the division into facets is a very weak
| form of structure.

* Peter VanDijck
| 
| Yes - if compared to ontologies. 

True. Admittedly I've brainwashed myself by working with this to the
exclusion of almost everything else for the past few years, so
ontologies have become the standard by which I judge all
classification systems. The result is that a lot of the things
companies do with thesauri, faceted classification, taxonomies, and
glossaries looks absolutely bizarre to me. (When I explain my view
they sometimes even agree. :)

| However, a simple division into facets makes it possible to make
| very powerful NAVIGATION widgets/systems (see Facetmap). 

Not too bad, admittedly. My formatted brain only sees a scaled down
ontology, but I admit it sucks less than many other things I've seen.
(Well, the interface could obviously be improved quite a lot, but I
should judge the approach rather than this specific implementation of
it.)

| I haven't seen any really good navigation widgets/systems for the
| more advanced ontology structures - it is usually just see-also and
| some ways of combining queries. 

I can't show you any of our really good demos, but here's two weak
examples that at least gives a hint of what can be done:
  <URL: http://www.ontopia.net/i18n/index.jsp >
  <URL: http://www.ontopia.net/operamap/index.jsp >

| My point is: we have some idea of how to build powerful navigation
| with Facets only, and adding a full blown ontology to that doesn't
| really add much to the navigational capabilities. (Please
| disagree/discuss if you will - I'm making this up as I go)

I've spent the past year and a half of my life creating tools to build
arbitrary navigational interfaces on top of ontologies, so it would be
difficult for me not to disagree. :-)

The whole point of applying the concept of topic maps to findability
is that it lets you build *very* rich navigational interfaces. From
any concept you can move in any number of dimensions. So the more
useful information you can connect to each node the more useful the
navigation becomes, and the key to the whole thing is that all
relationships are strongly typed.

Let's take one example:
  <URL: http://www.ontopia.net/i18n/script.jsp?id=latin-s >
 
Here you go to the Latin script. From this point you can go to the
type of script it belongs to, the category of scripts it belongs to,
the writing direction it uses, the script it was derived from, see the
entire family tree, the languages written in this script, to the
transcriptions from other scripts/languages to this one, as well as to
resources with more information.

Admittedly you have to be a grammatologist (or crazy) to care about
any of this, but don't see that faceted classification even allows you
to dream of doing stuff like this.

| Polyhierarchy is a feature for thesauri (from a classification point
| of view). The thing is: thesauri were never really structured for
| computer use. The thesauri structure is weak indeed, because they
| weren't designed by programmers. The whole programmer perspective
| (see what Travis is doing with structures in Facetmap, or see
| topicmaps) can add a lot of value to classification structures in
| general. I think thesauri in their old form are old school.

I very strongly agree. Where we differ, it seems, is on the question
of whether 
 
| Let's try:
| 
| - Geography (facet)
| -- Africa
| --- Morocco
| --- Nigeria
| ---- Kano (assuming Kano is in Nigeria)
| - Business areas (facet)
| -- oil services
| - departments
| -- oil surveying
| -- oil production
| - type of location
| -- office

This works, but is much weaker than an ontology. It also seems to
become unwieldy as the "ontology" is extended with more facets, and
you can't express the relationships between business areas and
departments, and so on...
 
| "Kano Field Location" would be classified in office, Kano and maybe
| some business areas and departments.

Actually, I don't think you could do that. "Kano Field Location" is
not a document, but a node in the tree. I think you'd either have to
put it under Kano and lose the organizational part, or put it under
oil surveying and lose the geography.
 
| Faceted hierarchies are definitely less powerful than ontologies, as
| in what they can express. But which system is easier to work
| with/implement? (this is not a rethoric question)

I would assume that faceted hierarchies are easier to implement, since
they are much simpler. Admittedly, that does count for something. I
doubt worse-is-better applies in this case, though.

As for which is easier to work with I am not sure how to approach that
question. Faceted hierarchies are less explicit, so they require less
effort to design, and at the same time accomplish less. For users I
think they provide a navigational interface that is less useful, but
for developers I would think they are easier because they do so much
less.

-- 
Lars Marius Garshol, Ontopian         <URL: http://www.ontopia.net >
GSM: +47 98 21 55 50                  <URL: http://www.garshol.priv.no >