The Role of Classification in Knowledge Representation and Discovery - 1

* Multiple perspectives. One of the most useful features of a faceted approach is that it allows entities to be viewed from a variety of perspectives--a feature that is lacking in hierarchies and trees. In a faceted analysis, it is possible to describe a dog as an animal, as a pet, as food, as a commodity, and ad infinitum, so long as the fundamental categories have been established with which to do this.

While the flexibility and pragmatic appeal of faceted classifications have made this a popular approach, there are some limitations in terms of knowledge representation and creation:

* Difficulty of establishing appropriate facets. The strength of a faceted classification lies in the fundamental categories, which should express the important attributes of the entities being classified. Without knowledge of the domain and of the potential users, this is often difficult to do. While it is possible to flexibly add entities, it is not a simple matter to add fundamental facets once the general classification is established.

* Lack of relationships among facets. Most faceted classifications do not do a good job of connecting the various facets in any meaningful way. Each facet functions as a separate kingdom, as it were, without much guidance as to how to put the parts together. For example, to facet analyze motion pictures by genre, country, director, film process, and so on, we would still have no insight as to the meaningful relationships of, say, a particular country and the popular film genre there or of a particular film process and the genres it supports. In terms of theorizing and model building, the faceted classification serves as a useful and multidimensional description but does not explicitly connect this description in an explanatory framework.

* Difficulty of visualization. A hierarchy or a tree, and especially a paradigm, can be visually displayed in such a way that the entities and their relationships are made evident. This is difficult to do for a faceted classification, especially if each facet is structured using a different internal logic. As a result, faceted schemes can only be viewed along one or two dimensions at a time, even though a more complex representation is actually incorporated into the descriptive strings. Thus it is difficult to see a vase in the context of other vases, of other Japanese artifacts, of other clay objects, of other raku objects, and so on, all at the same time.

Nevertheless, faceted schemes continue to flourish because we recognize that they allow at least some systematic way of viewing the world without the necessity for a mature and stable internal framework in which to view it. Information technology has promise for new ways of enabling multidimensional visualization and for developing computer-assisted ways of discovering patterns and anomalies that can possibly lead to new knowledge.

CLASSIFICATION AND KNOWLEDGE

There are many ways in which classification schemes and knowledge interact. Sometimes the interaction is so harmonious that the two remain linked for a long time. Sometimes knowledge changes and the classification must also change or knowledge changes and the classification is no longer adequate to the task. Sometimes the classification itself generates new knowledge. The following discussion is representative of ways in which knowledge and classifications mutually interact.