Talk:Faceted classification

From Wikipedia, the free encyclopedia

Contents 1 No link to MECE principle? 2 MECE is a bad idea 3 Major edit for clarity 4 More perspective needed

[edit] No link to MECE principle?

The [Knowledge Management Connection] article on Faceted classification mentions the MECE principle but that is not listed here. Any reason for that?

Incidentally Faceted classification is the theoretical foundations of the [Debian Package Tags] project. --16:48, 9 June 2006 (UTC)

[edit] MECE is a bad idea

Sometimes an item needs to be in more than one category. MECE is fine for libraries, where each book can appear on only one shelf. There are no such restrictions in the world of data.

[edit] Major edit for clarity

The existing technical explanation of faceted classification was (imo) too difficult to understand if you're not already in the field. I focused the article on navigation since that's how most people are going to encounter FC.

This version of the article could use a section on the technical side of the topic.

I added to the contrast with folksonomies the fact that folksonomies are bottom up and FC's are top-down.

I added a link to the Flamenco open source FC project.

dweinberger 20:43, 29 January 2007 (UTC)

[edit] More perspective needed

As someone who has been pushing the advantages of faceted classification for over 10 years, I was extremely happy to discover that a useful Wikipedia page had been created for the topic.

However, given the steadily growing use of faceted classification and navigation for online retrieval requirements, this looks like a good time to make this page an even better resource. And I hope that FC experts like Kathryn La Barre and Claudio Gnoli will chime in, too. (I'll send them pointers to this discussion.)

It's also important to note that the functions of faceted classification do not exist in isolation from other needs in the area of "enterprise knowledge management." And that's probably a good place to start this rambling commentary.

So, keeping in mind that I'm not a purist when it comes to applying FC in enterprises ...

First of all, it's useful to think of a faceted classification schema as multiple discrete "taxonomies" -- usually hierarchical arrangements of concepts based on whole-part or IS-A relationships. Each taxonomy is a single hierarchy representing a property (characteristic) of an object or an information unit. (OK, a facet can be completely flat, but that's not usually the case.)

Multiple-parent (polyhierarchical) relationships are not used in such taxonomies. And the separate taxonomies are chosen because they are typically semantically orthogonal to each other. Think People, Places, Location, Size, Color, etc.

In a well-designed interface for navigating information organized in this way, it is very easy to drill down to a small set of objects ... and just as easy to step "up" to higher levels of abstraction, because you are usually working with multiple taxonomies simultaneously. In fact, the ease and speed with which you can find objects organized in this way (that is, by properties rather than by commonly accepted name) is probably best expressed as "rapid information thinning" -- a phrase I first heard used by Giovanni Sacco.

I think it's important to note that multiple taxonomies also typically form the backbone of computer ontologies, and that the use of such ontologies within enterprises and industry associations is growing rapidly. The emergence of the Web Ontology Language (OWL) seems to have added impetus to that trend.

Although most developers of ontologies seem to focus on navigation and discovery of information within an ontology as best aided by graphic visualization of typed relationships among concepts, a faceted navigation interface to taxonomies in ontologies would be a natural complement to such visualization -- especially in large ontologies. I haven't seen this done yet, but maybe it's out there.

Another perspective on faceted navigation is that it is a navigational analog to retrieval of information from a relational database. It replaces construction of queries. This is especially true if you ignore the more "religious" FC viewpoint that you cannot select more than one concept from the same facet. Boolean logic applies equally well in both cases, IMHO.

Faceted classification is so natural that people are always re-inventing it in isolation. For example, take a look at the Wikipedia feature request for intersecting categories. It seems they're looking at representation of static intersections of categories, but the basic principle is the same. The discussion page has lots of interesting observations about basic categorization challenges -- most of which could probably be addressed by an experienced LIS expert with knowledge of FC.

And one last point: It was nice to see Flamenco become Open Source, but it appears that we may be facing some potentially nasty intellectual property wars, what with patent claims on various approaches to faceted navigation by leading vendors. (Thanks, Kathryn, for helping to point that out.) I hope it doesn't come to that. (And I can't comment objectively on this topic, because the company I work for has purchased and implemented one of the leading commercial FC-based products -- at my recommendation.) PhilMurray46 17:17, 1 February 2007 (UTC)