Previous: The Role of Ontology
Up: Ontology
Previous Page: The Role of Ontology
Next Page: Ontology-based intelligent information gathering

Weakly Structured Ontology

Unfortunately, development of ontologies is often a quite painstaking and time consuming task. Ontologies are often described in frame languages such as Ontolingua [1] and knowledge representation languages based on first-order predicate logic. We believe that the difficulty comes from the fact the these languages is computer oriented media and not human-oriented media. Since most of our knowledge is in human media such as natural language documents, we have to somehow translate human-oriented media into computer-oriented media. As human-oriented media is often ill-structured, i.e., ambiguous, indefinite, vague, unstructured, unorganized and inconsistent, we need a tremendous amount of efforts on translating ill-formed information into well-formed information.

We decided to make use of weakly structured ontologies which is developed from existing terminologies, thesauruses [3], and technical books [6]. Figure 3 shows a part of of an ontology about artificial intelligence.

Weakly structured ontologies have only one type of associative relation between terms. Conceptual relations such as concept-value, class-instance, superclass-subclass, part-whole are not explicitly distinguished in the weakly structured ontologies .

In the following experiments, we use the ontology built from the information science terminology which has about 4,500 terms.