Distinguishing Properties of Ontologies as We Know Them (10R4)
This is an experiment in collaboration and intellectual dialog. The question is: (10R6)
Can a group of people who care passionately about the definition of a word or concept collaborate effectively via a wiki on a description of what it is? (10R5)
Process (10RC)
The process for this wiki page is simple but it is important that we follow it to give the experiment a fair try. (10R7)
What: an attempt to characterize the notion of 'ontology' as it is used by practitioners (10RB)
Who: please only contribute to this page if you have built an ontology that has been shared over the Internet, or have used such an ontology in a computer program. (10R8)
Format: this page consists of a set of distinctions - properties of things that distinguish ontologies in this sense with other entities or artifacts that have been called ontologies. (10R9)
Rules of engagement: Under each distinction, add a bullet with text that makes your case, and sign it. Don't erase or modify other people's bullets. You may add a new distinction if desired, but please justify why it is essentially different than the distinctions already proposed. If you don't like the wording on a distinction, say that in your bullet commentary. (10RD)
Etiquette: Please write your comments offline in a word processor, then paste them in to this page when you are done. Otherwise we will have conflicts when multiple people try to edit the page at the same time. (10RP)
Distinctions (10RE)
Word Sense: Is there a word sense of "ontology (computer science)"? (10RF)
Should there be a separate entry in Wikipedia to discuss the notion of ontology as is used by the field of AI, the Semantic Web standards, and standards for data interoperability -- as compared with the subfield of philosophy from which the term originates? If you prefer another term for this distinction, such as applied ontology or computational ontology, why is it important for our field to use this term instead? (10RH)
- Yes, although the word 'ontology' clearly comes from philosophy, there is now a large body of literature and thousands of people doing work on entities that they call ontologies which are not exactly the same as what philosophical ontologists have created over the years. Ontology is now part of a technology stack, and is amenable to design methodologies like other digital artifacts. To make a new word sense is not to usurp or ignore the work of ontology as a field of philosophy; rather, it is to clarify a technical term so that practitioners can be clear when they use the term in presentations and publications. - TomGruber (10RY)
Representation or idea: Is an ontology (CS) a representation or an (idea | theory | view | abstraction) that is represented? (10RI)
Is an ontology (CS) an abstract idea -- like a Platonic ideal -- which exists independent of any form, or is it something that must be expressed in some form in order to be an ontology? Is the expression of an ontology like the implementation of an algorithm in the sense that people can talk about THE quicksort algorithm without referring to the programming language in which it is implemented? If an ontology must be expressed in some form, can it be translated into equivalent forms? If so, is there an important semantic difference between the two? (10RJ)
- An ontology (CS) is like an algorithm. It is an abstraction that can be stated in many different formalisms, which vary in precision and machine interpretability. However, it cannot exist outside of some representation. It is not the world or domain being represented. It is the representation of a simplified view of some world or domain. If an ontology exists independent of a representation then it is hard to understand what is meant by designing, testing, incorporating, copying, transmitting, or otherwise manipulating it in digital form. Therefore, it is important that we clarify that it is the representation, not the idea or theory in the conventional sense. - TomGruber (10RK)
- An ontology (CS) is like an artefact. It is a concrete form of an abstraction with specific limits predefined into its structure. When viewed as a logical theory, it references specific definitions of relations, which have particular characteristics. Not all inferences can be made from the relations, because their definitions segregate some inferences as being true and others as being false. An ontology is defined as much by what conclusions are allowed to be deduced as by which ones are not allowed. Distinctions that are not formalized allow for freedom in representation and nondeterministic states. There must be a balance between these two approaches in a particular ontology. Too rigid a system can be a stifling description and thus false to the reality represented. A system that is underspecified can be false when it does not provide logical support for all the inferences which can be derived about the reality represented. - DavidWhitten (10S4)
Conceptualization: What is the * represented or specified by an ontology (CS)? (10RL)
What is a good name and description for "the objects, concepts, and other entities that are assumed to exist in some area of interest and the relationships that hold among them"? (10RM)
- In the specification of conceptionalization definition, I chose this word because (1) it was in the AI literature -- in a textbook, and (2) it made the representation vs. idea distinction unavoidable present. As stated clearly in the definition, conceptualization was used as a shorthand for: "the objects, concepts, and other entities that are assumed to exist in some area of interest and the relationships that hold among them". Note that the expanded phrase is explicitly agnostic about whether things have to be concepts per se. I doubt that most people who don't know what an ontology is would be confused that it might be associated with "conceptualism". However, it is important to be technically coherent, so if there is a better name or phrase for this, add your bullets below. - TomGruber (10RZ)
Realism: Must all ontologies be about a One Shared Universal Reality? (10RN)
Or could ontologies represent abstractions of data with meaning that is independent of measurement or observation? (10RO)
- The question of whether terms in an ontology can only denote things in Reality is a great example of why computer and information science needs its own word sense. Data models have no requirement to be about any thing in the world. They are abstractions. Software Engineers may work with logical formalisms, but they often create models of imaginary worlds, and theories of worlds that are mutually inconsistent. Modern physics and epistemology acknowledges that genuine knowledge can be constructed, and modern brain research shows that humans do, in fact, construct their models of the world. So I only confusion when associating an ontology (CS) with Reality. The alternative position it to embrace ontologies (CS) as the artifact of deliverate design. In the realm of design, the 'right' design is always contingent on requirements and intended use. - TomGruber (10S0)
Formalism: Should the form (or medium) of an ontology be a necessary or sufficient condition or is it incidental? (10RQ)
Should all things we call ontology (CS) be encoded in some formalism with some specific properties? Is so, why distinguish those ontology-like entities that are represented in some other form? Should all things encoded in some class of representation be an ontology by definition? Is machine-readability a necessary condition? (10RR)
- The specification of conceptionalization definition text takes the stance that ontology (CS) is not defined by its form, but rather by its role as a technology for knowledge sharing and reuse. Formalism is an expedient choice, to satisfy requirements of precision, expressivity, or reasoning. However, it is true that most ontologies in the sense of the practitioner community are comprised of representational primitives such as (class | concept | unary predicate) and (relations | functions | nary-predicates) with definitional text and formal "constraints on the interpretation and well-formed use of these terms." It can be easier for a computer or information technology person to understand that an ontology is a "X made from Y's and Z's" than to define it in terms of its role in a technology stack. My personal opinion is that both approaches work. The difficulty with defining ontology by its form is that there is a wide variety of legitimate forms, and some would be excluded by describing the most common forms. However, if form is used to describe ontology, I find it useful to distinguish representational primitives from domain theories that use these primitives. And the problem is, there is no clear formal difference between the two. - TomGruber (10S1)
Specification: Is "ontology as specification" intrinsic to ontology (CS)? (10RS)
Or is the use of ontology as a specification only one of many possible uses? (10RT)
- If you have read this far I offer you a surprising reward. I think that associating ontology (CS) with specification is a statement of how it is used as a technology for knowledge sharing and reuse, rather than an intrinsic property of the beast. The rational for "specification" was to distinguish this sort of representation -- used as an interface layer and complexity hiding technique -- from the databases, knowledge bases, and assertions issued by agents that commit to the conceptualization. Just as many programs can implement an interface with different code, and many databases can instantiate a schema with different data, the ontology serves as this specification of the agreement among parties and not as the literal encoding of their data. So in this new set of distinctions, I would say that an ontology is a representation, plain and simple, and it is a representation of whatever this community decides to call a conceptualization. - TomGruber (10S2)
Vocabulary: Is it intrinsic that an ontology define vocabulary or representational primitives? (10RU)
What is the essential distinction between an ontology and any other representation or logical theory? (10RV)
- If asked for the most concise definition, I would say that an ontology defines a representational vocabulary. Then, if given a second sentence, I would explain what I mean by representational vocabulary: representations of the classes of entities, their properties and relationships in the domain, and constraints on their use. That is, an ontology is a representation, and of all the representations in the world, it has a special purpose of defining representational primitives for a domain of discourse. It offers a compact, carefully analyzed and socialized interface among parties in a knowledge sharing or reuse situation. This is sufficiently broad to include, in a meaningful way, controlled vocabularies and other sorts of taxonomies, database schemas, class hierarchies, files of OWL, and formal axiomatizations. - TomGruber (10S3)