Ontology dynamics and evolution through prototype revision

Abstract

An ontology in computer science is a formal explicit specification of the terms of an application domain, along with the relationships between these terms. An ontology provides a (structured) vocabulary which is the base for representing general knowledge. Ontologies have found numerous applications in Artificial Intelligence and the Semantic Web as well as in areas such as software engineering, bioinformatics and database systems. Recently, research in ontologies in Artificial Intelligence have focused on ontologies formalized through description logics (DL), which are (decidable) fragments of first order logics. Historically, an ontology in DL is divided into two components, the TBox to express concepts and their interrelationships, and ABox containing assertions about individuals and specific instances. More recently, there has been interest in the study of the dynamics of ontologies. An ontology tends to evolve, both in its domain information that can be corrected or refined, as in response to a change in the underlying domain. In a description logic, this change can come in two different forms: the background knowledge, traditionally stored in the TBox, can require modifications and specific data, traditionally stored on the ABox, can be modified. These changes may be treated with techniques of belief revision, properly adapted to the context. Among the potential tools that a priori seem more appropriate for implementation of dynamic ontologies is the model for revision of prototypes, developed jointly by the Responsible Researcher (Prof. Renata Wassermann) and Visiting Researcher (Prof. Eduardo Fermé).The model for revision of prototypes applies the dynamic of beliefs to the theory of prototypes developed by Eleanor Rosch in the 1970s, with the help of other experts in the field of cognitive psychology. In theory, people categorize concepts and individuals based on a prototype or ideal representation of the category. For example, the term dog is often characterized by skin, a tail and paws. When discussing or thinking about dogs, people think of classical stereotyped examples as collies or spaniels, because these represent the prototype. While a wolf or coyote can also meet the criteria fora dog, these animals are not prototype of a dog. According to the prototype theory, certain characteristics of a category have the same status, and thus examples presenting all or most of these characteristics become the prototype for that category. Items not sharing most of these characteristics may also belong to the category, but do not represent the prototype. The relations in ontologies can be thought of as the characteristics of the prototypes, while individuals may or may not belong to the class of prototypes, according to the properties they have. However, these classifications are never fixed and both prototypical properties as individuals change their characteristics: Thus, a Nineteenth century prototypical transport is probably not prototypical in the XXI Century, as well as the prototypical properties of a mobile phone have changed in recent years. Owning a formal model for these changes is then crucial. Recently, Fermé and Wassermann resumed work on the proposed model and developed an agenda that will begin during the visit. This includes the definition of prototypes in the TBox/ABox framework and: a) how to characterize the prototypical exemplars of a domain, b) how do you manage the dynamics of the concept of prototypical element when new elements are incorporated into the domain (prototypical or not), when pre-existing elements disappear, and how changing properties of prototypical objects affects the taxonomy. (AU)