The term “Ontology” was popularized in Computer Science by Thomas Gruber at the Stanford Knowledge Systems Lab (KSL). Gruber’s highly influential papers defined an ontology as “an explicit specification of a conceptualization.” (Gruber, 1992; Gruber 1993). Gruber cited a conceptualization as being “the objects and concepts, and other entities that are assumed to exist in some area of interest and the relationships that hold among them.” (Genesereth & Nilsson, 1987). The term “Ontology” has been used in computer science at least since (Neches, 1991), but is derived from philosophy where it defines a “systematic account of existence,” usually contrasted with “Epistemology.” Gruber’s work is firmly grounded in Knowledge Representation and Artificial Intelligence research going back to McCarthy and Hayes classical paper (McCarthy & Hayes, 1969). Gruber’s work also builds on frame systems (Minsky, 1975; Fikes and Kehler, 1985) which have their roots in Semantic Networks, pioneered by (Quillian, 1968) and popularized through the successful and widespread KL-ONE family (Brachman & Schmolze, 1985). One can argue that Gruber’s ontologies are structurally very close to previous work in frame-based knowledge representation systems. However, Gruber focused on the notion of knowledge sharing which was a popular topic at KSL around the same time, especially in the form of the Knowledge Interchange Format (KIF) (Genesereth, 1991). Ontologies have recently moved center stage in Computer Science as they are a major ingredient of the Semantic Web (Berners-Lee et al., 2001), the next generation of the World-Wide Web. Ontologies have also been used in Data Mining (see below) and in (database) schema integration.
TopIntroduction
Ontologies
The term “Ontology” was popularized in Computer Science by Thomas Gruber at the Stanford Knowledge Systems Lab (KSL). Gruber’s highly influential papers defined an ontology as “an explicit specification of a conceptualization.” (Gruber, 1992; Gruber 1993). Gruber cited a conceptualization as being “the objects and concepts, and other entities that are assumed to exist in some area of interest and the relationships that hold among them.” (Genesereth & Nilsson, 1987). The term “Ontology” has been used in computer science at least since (Neches, 1991), but is derived from philosophy where it defines a “systematic account of existence,” usually contrasted with “Epistemology.”
Gruber’s work is firmly grounded in Knowledge Representation and Artificial Intelligence research going back to McCarthy and Hayes classical paper (McCarthy & Hayes, 1969). Gruber’s work also builds on frame systems (Minsky, 1975; Fikes and Kehler, 1985) which have their roots in Semantic Networks, pioneered by (Quillian, 1968) and popularized through the successful and widespread KL-ONE family (Brachman & Schmolze, 1985). One can argue that Gruber’s ontologies are structurally very close to previous work in frame-based knowledge representation systems. However, Gruber focused on the notion of knowledge sharing which was a popular topic at KSL around the same time, especially in the form of the Knowledge Interchange Format (KIF) (Genesereth, 1991).
Ontologies have recently moved center stage in Computer Science as they are a major ingredient of the Semantic Web (Berners-Lee et al., 2001), the next generation of the World-Wide Web. Ontologies have also been used in Data Mining (see below) and in (database) schema integration.
Medical Terminologies
In the field of Medical Informatics a rich set of Medical Terminologies has been developed over the past twenty years. Many of these terminologies have as their backbone a taxonomy of concepts and IS-A (subclass) relationships. This IS-A hierarchy was pioneered in the semantic networks and frame systems mentioned above. With this structural commonality of ontologies and Medical Terminologies in mind, we will treat both kinds of knowledge representation systems together. Some of the largest existing ontologies have been developed in Medical Informatics, which makes this field especially interesting. For example, the Unified Medical Language System (UMLS; Humphreys et al., 1998) Metathesaurus contains information about over 1.5 million biomedical concepts and 7.2 million concept names.1
Background
The easiest way to understand ontologies is to look at them from a structural perspective. Quillian’s original semantic network was a computer data structure that mimicked a dictionary. In the KL-ONE implementation of a semantic network, the IS-A relationship took center stage. Thus, an ontology is (visually) a network of nodes (boxes) and links (arrows) connecting the nodes. Figure 1 shows a tiny excerpt of an ontology.2
Figure 1. Example of Medical Terminology applied to the treatment of bacterial pneumonia
The basic unit of knowledge in a terminology is a concept. Drug, Pneumonia, Antibiotic and Erythromycin are concepts. For each concept various kinds of attributes may be specified, e.g., name, ID number, synonyms and other alphanumeric attributes. These attributes provide additional information about the given concepts. In Figure 1, Dosage is an attribute.