Ontology building and population

An ontology is usually defined as a shared specification of a conceptualisation. For instance, for a particular domain, e.g. Natural Language Processing, to create an ontology, we need:

A common terminology for the concepts that are relevant in this domain.
A common set of relationships amongst these concepts. Some of the most common relationships in ontologies are:
- Hyperonymy, that relates a concept with other that is more general. For instance, computational linguist is a hyponym of linguist and or computer scientist, and, conversely, linguist is a hyperonym of computational linguist.
- Meronyny, the IS-A-PART-OF relationship.
- etc.
To define common axioms, or assertions, about those concepts and relations.

On the other hand, in an ontology it is common to distinguish between concepts, that represent sets of objects of interest with some shared properties, reason why it is useful to have a common name for them all; or instances (also called individuals), that denote examples or instances of the concepts. It is possible to argue that nothing is a concept or an instance, but that it depends on the point of view. However, in particular domains, and for most applications, it is generally possible to find a common agreement about what is a concept, and what is an instance. Concerning the automatic generation of ontologies, the following two tasks are very related:

Ontology building consists of structuring the concepts in an ontology, by means of relationships.
Ontology population consists of, given an existing ontology, populate it with instances of already existing concepts. Note that this task is similar to Information Extraction, in the cases in which one is asked to find inside a text entities and label them with categories taken from an ontology. In the area of semantic web, this task is usually called annotation.

Introduction

Publications

On-line demos

Some external links