The Internet Encyclopedia (Volume 3)

P1: 61

WL040C-197-Quin WL040/Bidgoli-Vol III-Ch-56 June 23, 2003 16:38 Char Count= 0

692 WEBCONTENTMANAGEMENT

Figure 4: Portion of the workforce ontology.

Once the data model stabilizes, it needs to be trans-

lated into a structure with more details. Such a structure

is sometimes called “knowledge structure” or “ontology”

(not in the sense of philosophical studies of the world ex-

isting, but in the sense of representing the real world).

Depending on the encoding language and software used,

the same knowledge structure may appear differently.

Figure 4 shows a portion of the ontology built from the

data model in Figure 3. Theprojectclass in the workforce

domain has a set of properties, each of which has name,

type, constraints (cardinality), and other facets if the prop-

erty type is symbol, class, or instance. For example, the

property “project sponsor” uses “instance” as the property

type. The class for property type “instance” is “Organiza-

tion,” from which instances will be extracted as the values

for project sponsor. Organization as a class has its own set

of properties. When it is used in another class (in this case,

the project sponsor property in the project class), the class

organization and its properties can be conveniently refer-

enced and reused, thus saving time from redeveloping the

same knowledge structure and improving consistency in

representing the same class in different locations within

the domain. The template properties provide a framework

for building schemas that will be used to design interac-

tive Web forms to capture data. Using the example in Fig-

ure 4, we can easily produce a database schema or an

XML schema that can be used to create XML instance

documents.

Controlled Vocabularies

Controlled vocabularies are tools for content classifica-

tion and indexing. Taxonomies, thesauri, classification

schemes, and glossaries are the forms of controlled vocab-

ulary frequently used in categorizing and indexing con-

tent. Taxonomy is a term borrowed from biology. It im-

plies three meanings according to WordNet (Princeton

University Cognitive Science Lab, 1998):

(1) a classification of organisms into groups based on sim-

ilarities of structure or origin;

(2) a study of the general principles of scientific classifi-

cation; and

(3) the practice of classifying plants and animals accord-

ing to their presumed natural relationships.

Taxonomies provide a method for structuring and classi-

fying things—living organisms, products, subjects—into a

series of hierarchical groups to make them easier to iden-

tify, study, or locate (Montague Institute, 2002). For digital

content, taxonomies play two roles: as category labels for

indexing content and as a tool for searching, browsing,

and navigation. Category labels may be assigned manu-

ally by human content/metadata creators or automatically

by using computer programs. The automatic approach

often involves categorizing sample documents manually

and then uses the manual result to train the system to

classify other documents automatically. Properly classi-

fied content will allow an organization to inventory and

monitor the Web content based on a structured under-

standing of user and community needs.

Taxonomies may be generated through automatically

extracting, analyzing, and categorizing structured and un-

structured content. The key question is how to extract dif-

ferent types of content and categorize them. Approaches

commonly used in taxonomy development include sta-

tistical and symbolic methods (Parsons & Wand, 1997).

The statistical (similarity-based) approach uses similar-

ity measures to classify real world objects, which can be

done automatically through computer programs. It is less

labor intensive and time consuming compared to man-

ual work, but also less likely to produce the organiza-

tional schemes the way people would. The symbolic (goal-

oriented or explanation-based) approach relies on domain

knowledge to determine the classes. Semantic networks

(Woods, 1991) are examples of the symbolic approach. No

matter which approach one uses, it is a good idea for tax-

onomy development to start with existing reference books

and encyclopedias in the domain. The benefit is that using

existing sources creates a familiar taxonomy that exhaus-

tively covers a topic.

In the example as shown in Figures 3 and 4, the tax-

onomy for workforce services used a symbolic approach.

By collecting concepts and terminology from workforce

documents and reference sources, the initial taxonomy

produced a three-level hierarchy of classes for the con-

cept Project:.

Project

Activities

Case management

Community audits

Continuous improvement

Cross training

Employer focused programs