RE: Special vs general schemes

Thanks, Aida for another thoughtful and challenging reply.

I apologize for the great delay in responding. I've been distracted by a
war, developing business opportunities, and a few other things.

First of all, I must remind you and others that I have no formal education
in library and information science, so one of the great values of these
exchanges for me is that it expose my assumptions in useful ways ... even
if, as often happens, it exposes my ignorance. The assumptions in these
discussions are *not* obvious in many cases. This fact of communication is
brought home to me every day in discussions with my wife. However, in our
case, the process of negotiating the exposure of assumptions is usually more
humorous than enlightening.


[Aida]
> I am picking a few points that don't sound right for me...
>
> [Phil]
> > Although the wine-categorization example is a simple one,
> > it is directly
> > connected to the objectives of a business devoted to the
> > consumption of
> > wine. The wine categorization facets reflect the
> > stakeholders' understanding
> > of what's important and essential in that domain.
>
> You are talking here about user aspect of information system. I
> considered this to be sorted out at the stage when one embarks
> to design vocabulary. The place of users in the design of
> information system is central and implicit. In effect it is so
> central that there is no need to mention it.

Andrew Otwell is right. ("It's dangerously easy to assume that everyone has
a clear understanding of users, what they need, and how they want
information, and then not mention it.")

But I would go beyond that. The fundamental characteristics of faceted
classification and the *systematic* (thanks for the emphasis on that term,
Prof.  Sacco) nature of faceted classification is such that it enables us to
redefine the traditional relationships among (and contributions of)
designers, developers, and users of a classification schema. I won't go into
that any further here.

> At this point one may think:  I have all this
> bottles of wine, producers, different profile of buyers etc.,
> tacit and implicit knowledge about processes involved which
> mounts up to thousands of concepts... where do I start? So one
> would like to believe that there is an advantage in having more
> general framework and methodology in that initial process of
> organizing this vocabulary as you are starting bottom up and
> you have handful of concepts you don't know what to do with.

There's no question that understanding library science, the historical
framework of faceted classification, and the methodologies used by experts
like yourself would be helpful in many cases. Perhaps every case. But I
wonder, too, whether they do not also dispose us to uncritical acceptance of
certain principles and practices.

What's more, I suspect that many developers of highly useful faceted
schemas -- and tools that can be used for developing faceted schemas -- have
had no knowledge at all of library science in general or faceted
classification in particular. (I know that is true in some cases.)

> Is
> there a reason why you can't take all the processes and then
> work within those towards the specific need of your system and
> your users/stakeholders?

> 13 faceted general categories we talk about is a methodology,
> or even, empty boxes you can put your schema in, while working
> on your vocabulary bottom-up. You should not be driven by it
> but rather supported by it.

I'm not sure there's a real distinction between being "driven by it" and
being "supported by it". At what point do you sacrifice what you have
learned from inference and examination on the altar of canonical general
categories?

> And you may need this if your
> stakeholders vocabulary is so big that you can't handle it as
> you can handle simple one like wine boxes. You can think of
> general categories as a classificationist's tool, mind map, but
> this should not affect the purpose of your vocabulary.

This is true (if I understand you correctly), at least in part. A specific
example in my experience is a facet for ACTIONS. ACTIONS have much greater
universality (if I can be allowed to use such a phrase) than, for example, a
facet for DISCIPLINES or MODES OF TRANSPORTATION.

>
> The real question is once you use them to build your specific
> vocabulary for your stakeholders - what is the point of having
> them?

Certainly you want to trim away what is not needed. Users of a simple
faceted wine-selection interface are not interested in "everything you can
say about the concept of wine." I'm not even sure that professors of
oenology want that! At all costs, the users of the resource should not be
forced to deal with abstractions and irrelevant information when they want
to deal with very concrete manifestations of those abstractions. (The
indexer, by contrast, needs to understand the abstractions in order to use
them effectively.)

> I think the point here is that CRG believes that everything you
> can say about concept of wine (i.e. that what you stakeholders
> need to be said) can be generally classified as entities, their
> kinds, their parts, properties, or processes, or agents,
> products, place or time.

[snip]

> If you did your job correctly when building faceted
> classification your cataloguer and indexer, not to mention your
> stakeholders, should not have a clue about processes, entities,
> properties etc. They should find system logical and easy to
> use, without being bothered with the intelligence behind it.

I certainly don't argue that point! Well, maybe I do, a bit. Stakeholders
should be *delighted* by the intelligence behind it. This is another area in
which Prof. Sacco's emphasis on the "systematic" characteristics of a good
interface to a faceted retrieval system is so significant. If the
information-seeker can *see* -- and *easily apply* -- the logic of
organization, then you have provided a great benefit to the information
seeker.

I'm *not* saying that it's easy to provide such "delight" as you move beyond
online product catalogs and other simple implementations. One thing that LIS
specialists and retrieval specialists have learned is that, for
larger/broader collections, information-seekers choose the simplest methods
possible for finding information ... even if those methods are extremely
inefficient.

This is why most of us hate TV remote controls with 92 buttons so much. Only
our children find them "delightful."

[snip]

> What do you mean when saying that orthogonal characteristics
> works superbly?

Simply, that "The metadata may be faceted, that is, composed of orthogonal
sets of categories." Source: "Flexible Search and Navigation using Faceted
Metadata," Jennifer English et al. As in your example below, color,
material, and purpose are mutually exclusive, orthogonal characteristics. (I
think "purpose" raises a lot of issues, but I don't want to address them
here.) The wine example and others mentioned in this list work remarkably
well.

>
> a) bad classification
>
> evening red dress
> morning dress
> green dress
> wool green dress
> silk dress
>
> here you can't make green evening dress as evening dress
> appears only as compound

I think that even us non-LIS folks clearly understand how awful such a
classification would be.

>
> b) good classification
>
> dresses by colour
> red
> green
> yellow
> ...
> dresses by material
> silk
> wool
> cotton
> ..
> dresses by purpose
> evening
> bathrobe
> sleeping garments
> ...
> Here you can build any compound you want
>
> I hope you mean more than this as this is an example from
> introductory lecture to classification systems in an
> undergraduate library school.

> This is where the story starts.
> One has to make sure that everybody understands this before we
> really start to talk about faceted classification on what you
> can do with it in knowledge organization.

Ah, here's another point where you can help dispel my ignorance.

I have seen many such examples of "good" classification. But I find them
confusing with reference to "faceted classification." I can recall similar
examples from "traditional" classification systems. I have referred to such
examples as "node-level faceting." (And I think the ANSI standard for
construction of monolingual thesauri also refers to this as faceting. Gotta
check that.) LIS expert Wendi Pohs of IBM did not hit me when I used that
term. She actually understood what I meant.

But what general principle or tradition in faceted classification (or
"traditional" classification, for that matter) tells the schema designer
here that  one should use the three distinguishing characteristics COLOR,
MATERIAL, and PURPOSE?

And, if there are such principles or traditions, how can they be
extrapolated to subdividing classes in rapidly developing new technology
domains?

Is "evening [dresses]" an example of a pre-coordinate synthesis of a
compound subclass? It seems to me that you *can't* build "any" compound you
want from the subclass "evening [dresses]." If this is a faceted schema, it
seems you've mixed  compound entries into a facet (CLOTHING, for example?).

For your example of "good classification," wouldn't it be sufficient in an
online retrieval system to keep COLOR and CLOTHING as separate facets? Can I
assume that this is referred to as a "post-coordinate" approach?

I apologize if I have misinterpreted your intent.

>
> >
> > Should we jam the knowledge organization developed in
> practical
> > implementations back into a broader conceptual framework of
> faceted
> > classification? Or should the broader framework adapt to
> > and be informed by
> > the real world experience of information seekers?
>
>
> On this list so far everyone tried to defend their own domain
> needs. Does this mean that we have to wait until a sufficient
> number of implementors moves jobs across different domains
> including not only business, commerce but education, culture,
> art and science  - and than wait to hear what did he/she found
> in common when building portal for art and science with portal
> for grocery shops or industry....

Um, er, yes.

OK, OK, maybe not exactly.

One of the assumptions here is "waiting." I think many specific faceted
classifications can be developed very quickly, and that we will immediately
see ways in which the lessons learned in isolated implementations can
benefit the practice as a whole. There are some things to be learned here
from "extreme programming." (See http://www.extremeprogramming.org.)

Another corollary assumption here, it seems to me, is that experts have to
examine and compare schemas to discern patterns, similar relationships,
semantic congruence of terms, etc -- in  painful detail. I contend that if
the initial facet analysis (and its evolution) are appropriately modeled,
then patterns of relationships and semantic congruence of concepts can be
detected and reconciled with computer assistance.

However, the ability to leverage these possibilities is heavily dependent on
developing a clear, explicit model for implementing faceted classification
schemas for business domains.

[snip]

Thanks again for such interesting, useful, and challenging posts!

    Phil

-------------------------------------------
"I have made this letter longer than usual, only because I have
not had the time to make it shorter."
-- Blaise Pascal, Lettres Provinciales, 1657

Phil Murray -- Chief Knowledge Architect
The Knowledge Management Connection | http://www.KMconnection.com
401-247-7899




------------------------ Yahoo! Groups Sponsor ---------------------~-->
Get a FREE REFINANCE QUOTE - click here!
http://us.click.yahoo.com/2CXtTB/ca0FAA/i5gGAA/0bmwlB/TM
---------------------------------------------------------------------~->

----------
Thanks for playing. To unsubscribe from this group, send an email to:
[email protected]

 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/