A different type of knowledge representation
Hierarchies
The most common way of representing knowledge in
classification schemes has been in hierarchies. The LC classification scheme
is mainly hierarchical; DDC, while containing aspects of faceting, is also
primarily hierarchical. Faceted analysis can be thought of as another way
of representing knowledge.
Kwasnik (1999)
compares different types of knowledge representation -- hierarchies, trees,
paradigms, and faceted analysis -- and assesses their various strengths
and weaknesses.
Hierarchies, Kwasnik notes, are useful for incorporating
knowledge about relationships between items, the detail with which items
can be described, and for economy of notation. But, hierarchies are problematic
for a number of reasons, including:
-
The existence of multiple hierarchies. While some
entities can be neatly slotted into one hierarchy, most cannot. Most entities
could be considered a member of several hierarchies. And hierarchies may
not have neat boundaries, but overlap.
-
Multiple and diverse criteria. Only so much information
can be put into a hierarchy, before it becomes too complex. We might place
the breed cocker spaniels within a hierarchy of dogs, mammals, and so forth;
but what if we want to consider wild, versus domesticated dogs?
-
Lack of complete and comprehensive knowledge. Hierarchies
are usually comprehensive, and show relationships of individual entities
within the overall structure. They are therefore based on complete knowledge
of the domain in advance. But, where the domain in question is little known,
or where relationships are unknown or ill-defined, hierarchies are difficult
to build.
To summarize, hierarchies are useful within well-known,
mature fields, containing well-defined, clear class boundaries; hierarchies
are useful where the relationships between entities within the hierarchy
are well developed.
Strengths of Faceted Analysis
Kwasnik discusses the strengths of faceted analysis.
These include:
-
Not requiring complete knowledge of the domain,
or the relationships between entities. It is therefore of particular use
in new and emerging fields.
-
Being hospitable, that is, able to accommodate new
entities easily. New technologies, objects, or concepts can be accommodated
more easily in a faceted classification.
-
Flexibility; Facets can be combined flexibly, resulting
in any number of new and interesting associations. The use of faceted analysis
is well suited for post-coordinate retrieval .
-
Not needing a strong theory to hold together; applying
faceted analysis to documents, would not require the many rules and tables
needed for both LC and DDC to function.
-
Allowing multiple perspectives on an entity. A dog
can be described as an animal, as a pet, as a commodity, or as food.
-
Allowing a variety of theoretical structures and
models; for instance, a faceted approach could incorporate more than one
perspective on how to look at an entity. A work of literature might be
treated in one facet according to a model of genres; in another, according
to a model for languages. This is particularly useful for dealing with
multidisciplinary works.
These aspects demonstrate some of the advantages
of faceted analysis over hierarchical forms of knowledge representation.
Other researchers have pointed out advantages
that faceted analysis has over other classification schemes.
Dahlberg
(1995) asserts that currently used classification schemes cannot keep
up with the growth of interdisciplinary knowledge; and that a classification
scheme must allow for combinations of concepts for multidisciplinary works.
And Foskett
(2000) asserts that facet analysis reflects a natural way of thinking,
and is therefore user-friendly to all.
|