Ontology (data-processing)

See also: Ontology

At the beginning, the ontology is a philosophical concept . Moreover, it was considered that the study of ontology was part of the Métaphysique (metaphysical general). In Data-processing and information science, a ontology is a structured whole of concepts making it possible to give a Sens to information. It is also a data model which represents a whole of concepts in a field and the relationship between these concepts. It is employed for to reason about the objects in this field.

The Concept S are organized in a graph whose Relation S can be:

  • of the relations Semantic S;
  • of the relations of Subsumption.
The main objective of an ontology is to model a whole of Connaissance S in a given field.

One can as say as ontologies are employed in the Artificial intelligence, the semantic Web, the software genius, it and like a form of representation of knowledge about a world or a certain part of this world. Ontologies generally describe:

  • Individuals: basic objects
  • Classes: units, collections, or types of objects
  • Attributes: properties, functionalities, characteristics or parameters which the objects can have and divide
  • Relations: the bonds which the objects can have between them
  • Événements: changes undergone by attributes or relations

Definitions of an ontology

Abstract approach

The etymology returns to the “theory of the existence”, i.e. the theory which tries to explain the concepts which exist in the world and how these concepts are imbricated and organize themselves to give Sens.

Contrary to the human being, the Connaissance for a Computing system is limited to the knowledge which it can Représenter .

At the human being, the Connaissance S Représentable S (i.e. the Universe of the speech ) are supplemented by knowledge not exprimables (Sensation S, Perception S, feelings not verbalisables, knowledge Inconsciente S, etc). These nonrepresentable elements however take part in the processes of Raisonnement and Décision, which is cognitive Processus S in Knowledge management. The cognitive performances of a data-processing agent thus partly will rest on the field of the Représentation S to which it will have access, i.e. concretely with the field of the Représentation S which will have been formalized .

Data-processing ontologies are tools which precisely make it possible to represent a corpus of knowledge in a form usable by a Ordinateur.

One of the definitions of the ontology which makes authority is that of Gruber (cf references):

; An ontology is the specification of a conceptualization of a field of knowledge.

This definition is based on two dimensions:

  • an ontology is the conceptualization of a field, i.e. a choice as for the manner of describing a field.
  • It is in addition the Spécification of this conceptualization, i.e. its formal description .

It is a base of formalization of knowledge. It is located at a certain level of abstraction and in a particular context.

It is also a Représentation of a shared and consensual conceptualization, in a particular field and towards a common objective. It classifies of categories the relations between the concepts.

Criteria of evaluation of an ontology

According to Gruber, five criteria make it possible to highlight important aspects of an ontology:

  • the clearness : The definition of a concept must make pass the desired direction of the term, in manner as objectifies as possible (independent of the context). A definition must moreover be complete (i.e. defined by conditions at the same time necessary and sufficient) and documented in natural language.

  • the coherence : Nothing which cannot be inféré of ontology must enter in contradiction with the definitions of the concepts (including those which are expressed in natural language).
  • the extensibility : The extensions which could be added to ontology must be anticipated. It must be possible to add new concepts without having to touch with the foundations of ontology.
  • a minimal deformation of encoding : A deformation of encoding takes place when the specification influences the conceptualization (a given concept can be simpler to define in a certain way for a language of ontology given, although this definition does not correspond exactly to the initial direction). These deformations must be avoided as much as possible.
  • a minimal ontological engagement : The goal of an ontology is to define a vocabulary to describe a field, if possible in manner supplements ; neither more, nor less. Contrary to the knowledge bases for example, one does not await ontology that it is able systematically to provide an answer to an arbitrary question about the field. an ontology is the weakest theory covering a field; it defines only the terms necessary to share knowledge related to this field.

Operational approach

Parallel to this rather theoretical definition of what an ontology represents, another definition, more operational, can be formulated as follows:

; An ontology is a semantic network which gathers a whole of concepts describing a field completely. These concepts are related the ones to the others by taxonomic relations (hierarchisation of the concepts) on the one hand, and semantic on the other hand.

This definition makes possible the writing of languages intended to implement ontologies.

To build an ontology, one has at least three of these concepts:

  1. Determination of the passive or active agents.
  2. Their functional and contextual conditions.
  3. Their possible transformations towards limited objectives.
To model an ontology, these tools will be used:
  1. To refine the vocabularies and adjacent concepts.
  2. To break up of categories and others topics.
  3. Prédiquer in order to know the adjacent transformations and to direct towards the internal objectives.
  4. To relativize in order to include concepts.
  5. To harmonize in order to reduce to completely distinct bases.
  6. Instancier in order to reproduce the whole of a " branche" towards another ontology.

Ontologies in practice

Example of ontologies

For example, to describe the concepts entering concerned the design of electronic charts, one could define ontology (simplified here) following:

  • a electronic chart is a whole of component ,
  • a component can be either a condensing , or a resistance , or a chip ,
  • a chip can be or a storage unit , or a calculating unit ,
  • a electronic chart which contains a calculating unit contains also at least a storage unit .

Languages for ontologies

The language of specification is the central element on which ontology rests.

The majority of these languages base on the Logique first order, and thus represent knowledge in the form of assertion (subject, predicate, object). Among the formalisms most employed basing itself on the logic of the predicates, one finds languages like N3 or N-Triple.

One can also evoke the language ***.

In addition, within the framework of its work on the semantic Web, W3C set up in 2002 an work group dedicated to the development of standard languages to model ontologies usable and exchangeable on the Web. Taking as a starting point preceding languages like DAML+OIL and by the theoretical bases of the logical of description, this group published in 2004 a recommendation defining the language OWL ( Web Ontology Language ), founded on the standard RDF and by specifying a syntax XML. More expressive than its predecessor RDFS, OWL quickly took a dominating place in the landscape of ontologies and is from now on, de facto, the standard more used.

Tools to work with ontologies

The following editors of ontology are free and downloadable

  • Protégé more is known and more used editors of ontology. Open-source, developed by the University of Stanford, it evolved/moved since its first versions (Protected-2000) to integrate from 2003 the standards of the semantic Web and in particular OWL. It offers many optional components: arguers, graphical interfaces.
  • SWOOP is an editor of ontology developed by the University of Maryland within the framework of project MINDSWAP. Contrary to Protégé, it was developed in a native way on the standards RDF and OWL, which it deals with in their various syntaxes (not only XML). It is an application lighter than Protégé, less advanced in term of interface, but which integrates also tools of reasoning.
  • KMgen is an editor of ontology for language km (km: The Knowledge Machine).

With the emergence of the market of technologies of the semantic Web, one can note the appearance since 2005 of software tools proposed by commercial editors. One can quote:

  • SemanticWorks belongs to the continuation of tools XML developed by Altova. It supports language OWL through its syntax XML.
  • TopBraid Composer is developed by TopQuadrant. Its interface and its functionalities resemble much those of Protected (the principal developer of TopBraid being the old developer of extensions OWL of Protected).

There exist in addition tools Informatique S making it possible to build an ontology starting from a corpus of texts. These tools traverse the text in the search of terms recurring or defined by the user, then analyze the way in which these terms are connected in the text (by grammar, and the concepts that they recover and whose definition can be found in a lexicon provided by the user). The result is an ontology which represents total knowledge that the corpus of text contains on the scope of application that it covers. The WordNet project (see the bonds) is the most important example.

Normative approach

In Europe, the Norme which is the subject currently of an special attention is a standard which in particular makes it possible to describe ontologies on the immaterial Cultural heritage (Bibliothèque S, Musée S and Archive S,…). Its exact references are:

ISO 21127: ontologies necessary to the description of the data concerning the cultural heritage

See too

Random links:Causa perdida | Caussou | Cambrien (Celtic language) | Luminous Alfe | Lake Espejo | Chief candidate of the American standards | Enzo_Ferrari