In the philosophical sense the term ontology is used as to mean: "The science or study of being; that department of metaphysics which relates to the being or essence of things, or to being in the abstract" [Oxford English Dictionary]. But the term has surfaced in the computer science community, especially in knowledge engineering, natural language processing, cooperative information systems, intelligent information integration, and knowledge management [Fensel, 2000], and its purpose is there described to offering: a shared and common understanding of some domain the can be communicated across people and application systems [Fensel, 2000] or enabling knowledge sharing [Gruber, 2000]. That is precisely what the Semantic Web needs. Thus, the concept of ontologies sounds reasonable to use. Though, it is important to always keep in mind that the Web is a universal and highly heterogeneous environment.

There exist numerous descriptions and definitions of what an ontology is as used in the computer science community: "a shared and common understanding of a domain that can be communicated between people and heterogeneous and distributed systems." [Fensel, 2000]; "a computer model of some portion of the world." [Huhns & Singh, 1997]; "a specification of a conceptualization." [Gruber, 2000]. Although the one by Gruber is most commonly used, and a bit further explained by him, a need for a more formal definition is needed to avoid confusions and to fully understand what constitutes an ontology.

Definition of ontology and related concepts

The definition of ontology that is used throughout is the one presented by [Guarino, 1998] that extends Gruber's definition, and it is as follows:

"An ontology is a logical theory accounting for the intended meaning of a formal vocabulary, i.e. its ontological commitment to a particular conceptualization of the world. The intended models of a logical language using such a vocabulary are constrained by its ontological commitment. An ontology indirectly reflects this commitment (and the underlying conceptualization) by approximating these intended models."

This definition is also illustrated in figure 6.7 below, reproduced from [Guarino, 1998].


Figure 6.7

The definition clearly states very important properties that an ontology has. Each word used (for a relation or property) must explicitly declare its ontological commitment. In the case of the Semantic Web this means that we have to use URIs to achieve precision and to show its ontological commitment in the case of properties. The ontological commitment governs the use and meaning of the word, and as a consequence the applications' use of it. Further, an ontology is only a formal approximation of the world. Another important property that Guarino further states is that the ontology is language-dependent while the conceptualization is language-independent. This is in my opinion extremely important to notice. Two different ontologies might share the same conceptualization of a thing, e.g. the concept of price, but might map different words from their vocabulary to that concept. This means that it will become possible to achieve interoperability between different ontologies if the conceptualization is the same in some sense. But it also means that if two applications use the same words it does not mean that they map to the same conceptualization.

Every application could be considered to have its own conceptualization, and if two applications are going to be able to exchange "meaning" they have to have overlapping conceptualizations. As [Guarino, 1998] says, two systems A and B using the same language L can communicate only if the set of intended models IA(L) and IB(L) associated to their conceptualizations overlap. This is depicted in the left side of the figure 6.8 below:


Figure 6.8

On the right side of figure 6.8 it is showed that even if the intended models of two different ontologies do not overlap, the axiomatizations (set of axioms) might overlap. An example of this if one ontology describes human - man relations and states a number of axioms, this entire set of axioms might be the same as another ontology that describes vehicle - car relations. This does not mean that the intended models overlap, and the applications committing to those ontologies cannot communicate (meaningfully).

But there exist a few notable limitations: an computer application can't generally decide if two concepts overlap, this has to be decided by humans and explicitly described by a human expert by explicitly express equal-statements that connects word with different ontological commitments. This is important to consider.

There exists a small confusion, which is important to bring order into, and that is the use and scope of ontologies. In [Studer et al., 2000] the authors state that: in order to reuse an ontology as much as possible, ontologies should be small modules with a high internal coherence and a limited amount of interaction between modules. This is in my opinion wrong. (Although is it true considering software reuse.) Every concept and every relation and every property should have a URI. Thus, using its URIs enables the reuse of an ontology. And most importantly, the true semantical interoperability will be possible IF ontologies are highly interconnected. The Semantic Web will reach its true potential if different ontologies that share concepts are interconnected. As [Hendler, 2001] says there is most probably not going to be one big ontology that everyone is using, rather it will be a big set of different possibly interconnected smaller ontologies. (Although, I would rather use highly instead of possibly interconnected.) Of course, an ontology in the biological domain will be very different from an ontology in the philosophical domain, but they might have concepts that overlap (e.g. the author of a document) thus connecting them.

How does this ontology concept relate to the language at level 1? (Not that this is an extract from my thesis, thus parts are missing) All properties and classes of objects should be described by an ontology. An ontology is a bit simplified a set of interconnected properties and a set of logical properties. But in order to create this sort of property-networks we need a set of atom primitives, and set of standard properties that constitute the building blocks for connecting properties. Using these constructs to build a network of properties and then use this network to map properties of statements onto is what using an ontology is all about. Also, if we construct a property-network partial understanding is possible. If a set of statements uses a set of properties (i.e. a set of nodes in the properties network) partial understanding could be possible if another application uses a set of properties that are "near by" or closely related. By following routes in the network there might be a common base property that facilitates the partial understanding. This also means that isolated networks of properties are of lesser use than a big combined network (i.e. do not invent ad hoc properties without first looking for existing ones). What is generally needed in order to create an ontology is the ability to model classes of meanings and their relations [Huhns & Sing, 1997]. Thus, we need the possibility to create a kind of dictionary that defines the terms that will be used in statements and gives specific meaning to them.

[Fensel, 2000] Fensel D, 2000: Ontologies: Silver Bullet for Knowledge Management and Electronic Commerce, URL: http://www.cs.vu.nl/%7Edieter/ftp/paper/silverbullet.pdf
[Gruber, 2000] Gruber T, 2000: What is an Ontology?, URL: http://www-ksl.stanford.edu/kst/what-is-an-ontology.html
[Guarino, 1998] Guarino N, 1998: Formal Ontology and Information Systems, Amended version of a paper appeared in N. Guarino (ed.), Formal Ontology in Information Systems, Proceedings of FOIS'98, Trento, Italy, 6-8 June 1998, Amsterdam, IOS Press, pp. 3-15 URL: http://www.ladseb.pd.cnr.it/infor/Ontology/Papers/FOIS98.pdf
[Humns & Singh, 1997] Humns N & Singh M: Ontologies for Agents, IEEE internet Computing, Nov - Dec 1997, http://computer.org/internet
[Studer et al., 2000] Studer R, Decker S, Staab S, 2000: Situation and Perspective of Knowledge Engineering http://www.db.stanford.edu/˜stefan/paper/2000/ios_2000.pdf

Created with AmayaValid HTML