[Sigia-l] Creating Taxonomies/FC Systems from Data Models

[Doug Howell (IT)]

Hi, all, This is my first post. I've been reading this list off and on with interest for several months. I've much appreciated the experiences and thoughts you all have shared from your wonderfully varied points of view. 

My background is in music (degree in Music Composition). I've done quite a bit of personal and commercial composing, arranging and performing. I got hooked on computers (Macs) through music applications (Mark of the Unicorn's Digital Performer is still the most elegant, complex, fun-to-use program I've ever experienced), and learned more and more about them as I went along (MIDI forced me to learn hexidecimal!). The digital-music connection led to a musical dream job where I got to compose music full time for money (I had the chance to compose/perform soundtracks to about a dozen children's CD-ROMs). Unfortunately, about then the CD-ROM-bomb hit. Technical writing, which I'd done on the side, now became my full-time gig. Luckily, the VP that hired me saw this history as an advantage. After three years as tech writer, I became an Information Architect (not specifically Web related).

I used to wake up in the middle of the night with musical ideas---still do occasionally---but the other night I woke up thinking about faceted classifications. (That's how twisted my thinking has become.) Anyway, here's my question.

Has anyone pondered the possibilities of using data models to define taxonomies or faceted classification schemes? I've been very interested in FC, but the more I read about it here, the more daunting the prospect of actually developing such a system seemed. The nocturnal revelation was that maybe a lot of the work had _already been done_---in the form of our existing enterprise data models. These entity relation diagrams (ERDs) had been developed over a long period of time by a dedicated IT team for the purpose of establishing an enterprise-wide data dictionary that was not only representative of current data needs but future needs. It was to become the basis of our future application data structures.

Some related thoughts:

Scope - A critical step would be to figure out just how much and what portion of the data model to use as a foundation. Several have commented that the most successful FC schemes are clearly focused ones.

Granularity - Facets/terms could be created from either entities or attributes. How would the type of attribute/entity (strong, weak, associative) affect this?

Names - Most entity/attribute names could not be used as is. User labels would have to be assigned to make them human-readable. These labels could be stored in the data dictionary.

Relationships - To what extent do ERD relationships (identifying, unidentifying, cardinality, etc.) define or imply hierarchies or other thesaurus-term relationships.

Seems like ERDs could be very fertile ground for growing classification systems, not only because they provide a major resource for already-defined terms, but also because those terms are relatively stable and become the data definitions used by applications. 

Thanks,
Doug

Doug Howell
Information Architect
Borders Group Inc.
dhowell nospamat bordersgroupinc.com