What is ontology?
The American Heritage Dictionary (4th ed.) provides the following definition: “The branch of metaphysics that deals with the nature of being.” In short, ontology is the centuries-old branch of philosophy that has as its subject the unchanging features of the universe.
Slightly Longer Answer:
Barry Smith provides the following definition:
Ontology is the science of what is, of the kinds and structures of objects, properties, events, processes and relations in every area of reality. For an information system, an ontology is a representation of some pre-existing domain of reality which:
(1) reflects the properties of the objects within its domain in such a way that there obtains a systematic correlation between reality and the representation itself
(2) is intelligible to a domain expert
(3) is formalized in a way that allows it to support automatic information processing
An ontology in this sense is a thing made by a scientist or other domain expert. This thing is a formal theory, which accurately recapitulates the domain in light of the kinds of entities contained within it; that is, it is not ad-hoc, but in conformance with the world. Thus, an ontology is a true-to-the-world representation of its domain (SIGINT, calcium-regulated pathways, aircraft parts, etc.). This stands in contrast to the more popular usages by many in the fields of information and computer science, which see an ontology as merely an ad-hoc model built for some specific purpose.
The Ontology Works IODE is software designed to produce ontologies – true-to-the-world information models. In the IODE, the ontologies are actually extensions of an existing body of knowledge, which is the fruit of the discipline of ontology.
What about Gruber’s definition?
Tom Gruber says that “an ontology is a specification of a conceptualization.” His definition is incomplete and vague.
Gruber’s definition allows for the reduction of the meaning of “an ontology” to “a model”, where what is being modeled are the concepts or ideas people have in their minds. This reductive error has its roots in the recent tendency to use the word “ontology” to mean little more than a controlled vocabulary with hierarchical organization. Gruber’s definition fails to stipulate that if the theory or model in question is to be useful (for the integration and processing of real-world data and information), it must be about the universe and not merely about the concepts in people’s heads.* Such modeling is a perfectly reputable activity, but it belongs not to ontology but to linguistics, or psychology, or to another venerable field of philosophy: epistemology. Moreover, Gruber’s definition does not account for the existence of good and bad ontologies, where the former are the more accurate descriptions of the relevant domain of real-world entities.
Ontology is not about peoples’ conceptions or interpretations, but about the world.
What is the W3C’s Web Ontology Language (OWL)?
Computer-understandable ontologies are represented in logical languages, such as the W3C OWL (Web Ontology Language). However, logical languages are only a means to express content; they are themselves devoid of informational content. This situation is much like how the natural language English relates to information expressed in English. It is the information being imparted in the words that drives how the individual words are selected and sequenced into sentences. It’s not the language (or logic) that makes the difference, but how you use it. Ontology is one way to use language and logic more effectively.
The W3C’s Web Ontology Language (OWL) is but a logic, you are on your own when it comes to saying something with it and of ensuring the accuracy of your assertions.
What do ontology-based information systems have to do with Artificial Intelligence?
AI systems deal with applied reasoning tasks and the solution of problems, while ontology deals with the representation and retrieval of data. Ontologies can be used to support AI systems, by providing a deeper and more robust representation of the domain on which one wishes to reason and solve problems. For example, an AI system could be used to operate a series of cranes at a worksite for the purpose of moving rocks from a quarry to waiting trucks. An ontology could be used to help the AI recognize and differentiate different kinds and behaviors of rocks and trucks.
Not to be confused with AI systems that allow for inductive or speculative reasoning, Ontology Works software implements strictly deductive reasoning. It does not involve fuzzy logic, probability-based logic, or any reasoning that attempts to simulate consciousness.
*To read about what happens when people model the ideas and concepts in their minds, see Wuesteria.