Making Faceted Classification more acceptable on the Web: A comparison of Faceted Classification and ontologies

Authors


Abstract

Both Faceted Classification and ontologies offer a consistent structure to organize web resources effectively. This paper compares Faceted Classification and ontologies to introduce the concept of making Faceted Classification work more efficiently on the Web. It concludes that if complemented by ontologies, Faceted Classification could be more usefully accepted on the Web with increased conceptual expressiveness among facets.

Introduction

Faceted Classification has become of interest in organizing resources on the Web, because it does not fix the relationships among classes. Thus, Faceted Classification is more effective than traditional classifications for describing the kind of diverse and multidisciplinary subjects that are commonly found on the Web. However, Faceted Classification lacks an ability to express relationships among facets (Kwasnick, 1999). Unlike Faceted Classification, ontologies can formally express such relationships among fundamental concepts. The author notes that both Faceted Classification and ontologies have something in common with regard to their purpose: they offer a consistent structure to organize web resources effectively. The author discusses ontologies, their potential use in Faceted Classification for organizing web resources and their semantic (formal) aspects. In this paper, the author focuses on one ontology, the Resource Description Framework (RDF) ontology which has a strong similarity with Faceted Classification in terms of theory, mechanism, and hierarchy.

Problem with Faceted Classification

Faceted Classification, developed by Ranganathan and based on the Colon Classification scheme, is an “analytico-synthetic” scheme that derives from two processes: analysis and synthesis. Ranganathan defines these: (1) analysis as the process of breaking subjects into basic concepts and (2) synthesis as the process of combining those concepts into descriptions for subjects (Foskett, 1982, p. 390; Taylor, 1992, pp. 320-321).

However, Kwasnick (1999) notes that Faceted Classifications do not explicitly connect the various facets “in any meaningful way” (p. 20). Ontologies, on the other hand, will support this because they allow explicit and formal expression of concepts and relationships between concepts.

RDF vs. Faceted Classification

1. Theory

The author points out that the RDF model theory corresponds to the Faceted Classification theory. One first needs to look back at Ranganathan's Faceted Classification theory. Ranganathan asserted that “it is the duty of documentalists to spread the multidimensional universe of knowledge along one line” (quoted in Tinker et al. 1999). This theory means that multi-dimensional space should be mapped to one-dimensional space, a single standardized representation of the “universe of knowledge.” Tennis (2004) compares Ranganathan's theory with the RDF model theory. According to Tennis, in RDF, “N-dimensional space must be converted into machine readable relationships” (2004, p.2). The author argues that RDF defines the machine-understandable semantics of N-dimensional resources by the “triple model.” RDF is fundamentally a “framework for metadata” (Lassila, 1999). RDF represents metadata as models that consist of a collection of statements about resources. Statements consist of a subject, a predicate (or property), and an object that form the RDF triple model (Resource, Property, and Value) Figure 1.

Figure 1.

Simple statement graph template (Lassila, 1999)

RDF expresses simple but formalized assertions about web resources: resources (identified by URI) have properties (attributes of a resource or binary relationships between resources) with certain values (e.g. string or URI). For example, let us consider the following statement: Julie Aigner-clark” is the author of the resource My First Book of Colors (http://www.babyeinstein.com/MyFirstBookOfColors).

This statement can be represented as an RDF triple Figure 2.

Figure 2.

RDF graph representation of the statement

Also, the statement can be formally expressed below in RDF by the following schema,

original image

As seen in the examples above, in the RDF model, all resources are mapped onto the property-value relationships (Resource-Property-Value), which shows us that the RDF follows Ranganathan's theory.

2. Mechanism

The author asserts that the mechanism provided by the RDF Schema (RDF's vocabulary description language) describes “facets.” The RDF Schema (RDFS) describes properties in terms of the classes of resource to which they apply: domain and range of values. At this point, the author argues that facets are basically the properties' “domains,” and each facet provides a view of a unique aspect of the domain.

Figure 3.

Faceted Classification of Library Material

For example, given a Faceted Classification of library materials based on characteristics of physical form, publishers, and symbolical form, etc. Figure 3 (Vickery, 1960), one can define the “Author” property to have a domain of “Book” and a range of “String.” The following example shows how the RDFS describes a facet by specifying a domain (e.g., “Book”) to which the property (“Author”) may be applied:

original image

3. Hierarchy

The author remarks that both RDF and Faceted Classification share the same idea of “class hierarchy.” In the class hierarchy, a subclass inherits properties from its super class. Although each facet in a Faceted Classification is not necessarily hierarchical (general-specific, whole-part, etc.) (Knowledge Management Connection, 2007), the author asserts that Faceted Classification relies on the “class hierarchy” of facets, because under each facet (superclass), subfacets (subclasses) are listed, and then one subfacet splits into smaller subfacets. Priss and Jacob (1999) also note the concept of hierarchy while defining a “facet hierarchy” as “a set of facets with a subfacet/facet relation which forms an ordered set” (p. 6). Thus, the facet hierarchy implies a transitive inheritance of properties through facets. For instance, for one facet, such as “Book” (in Figure 3), within in the hierarchy, one can transitively infer “Codex” and “Folded.” The RDFS supports the same idea of class hierarchy as Faceted Classification. The RDFS also provides a transitive inheritance of classes and properties by means of abstract mechanisms: rdfs:subClassOf (refers a subclass to its superclass) or rdfs:subPropertyOf (refers a subproperty to its superproperty). For instance, if class A is a subclass of B, and B is a subclass of C, then A is also implicitly a subclass of C (Brickley, 2004). That is, in RDFS, the following:

original image

implies:

original image

Concluding Remarks

Concerning an aspect of the continuing relevance of Faceted Classification to the organization of resources on the Web, it can be summarized that Faceted Classification could be more usefully accepted if complemented by ontologies' formal descriptions of relationships among facets. To improve the ability of Faceted Classification to better support the description of resources on the Web, the next step should be the development of a formal semantics for a knowledge representation model that supports the description of more complex concepts and semantic relationships.

Ancillary