On Belief Change and Ontology Evolution

Giorgos Flouris

University of Crete,
Department of Computer Science
P.O. Box 2208, Heraklion, Crete, 
GR-71409 Greece
fgeo@csd.uoc.gr

Doctoral Dissertation
Heraklion, February 2006

Abstract
One of the crucial tasks to be performed towards the realization of 
the vision of the Semantic Web is the encoding of human knowledge in 
special structures (ontologies), using formal representation 
languages. Simply creating an ontology is not enough though; 
knowledge needs to be updated as well. There are several reasons why 
knowledge in an ontology would need to change: changes could be 
initiated because of a change in the world being modeled; the users’ 
needs may change, requiring a different conceptualization; knowledge 
previously unknown, classified or otherwise unavailable may become 
known; or a design flaw may have been noticed in the original 
conceptualization.
In all these cases, the ontological representation would need to be 
modified so as to form a more accurate conceptualization of the 
domain. In this dissertation, we will use the generic term ontology 
change to refer to the process of modifying an ontology in response 
to any given need. This generic term engulfs several different, but 
closely related research fields, including ontology evolution, 
alignment, merging, mapping and others. 
The first part of this dissertation deals with all these disciplines, 
performing a shallow, yet broad review of recent work in those areas 
and proposing a certain terminology in a field plagued by the use of 
underspecified and confusing terms. This will allow us to draw a fine 
line between these closely related and often overlapping research 
efforts.
The second part of this dissertation focuses on ontology evolution, 
which refers to the process of modifying an ontology in response to a 
change in the domain or its conceptualization. We will argue that 
current research on ontology evolution is misguided and leads to 
several problems, mostly related to the heavy dependency of the 
process on human participation and the lack of adequate 
formalizations. The key idea of our proposal towards resolving these 
deficiencies is the use of intuitions from the closely related field 
of belief change, which deals with the updating of knowledge stored 
in Knowledge Bases. 
We will study the connections between the two fields and propose a 
formalization for ontology evolution that facilitates the migration 
of belief change techniques, tools, ideas, results and intuitions to 
the problem of ontology evolution. The solutions proposed in the 
belief change literature generally have all the properties desirable 
for an ontology evolution algorithm (automatic determination and 
implementation of changes, formal character and well-studied 
theoretical properties), so they could be of use in the ontology 
evolution field. This way, ontology evolution researchers will not 
have to “reinvent the wheel” for problems whose counterparts have 
already been studied in the belief change literature and will be 
protected from potential pitfalls and dead ends.
It is obvious that exhaustively pursuing the migration of belief 
change techniques to the ontology evolution field is a long-term 
goal, not achievable within the framework and limitations of a PhD 
dissertation. For this reason, this part of our work will only 
provide abstract ideas towards this direction. Our aim at this point 
will be merely to provide the necessary theoretical foundations and 
connections upon which future research could be based, hopefully 
leading to more concrete solutions to the problem of ontology 
evolution.
Following this abstract study, our research will focus on a certain 
belief change theory, namely the AGM theory, which is the most 
influential approach in the field of belief change. This study 
constitutes the third part of this thesis and its aim is to provide 
an exemplary application of the general research path we are 
proposing, as well as to show its feasibility.
Our study of the AGM theory will be initiated by an attempt to 
generalize it so as to be applicable to logics that are used for 
ontological representation. The AGM theory, like most belief change 
approaches, focuses on classical logic, disallowing, among others, 
the main formalisms used in ontologies, like Description Logics (DLs) 
and the Web Ontology Language (OWL). Our generalization will cover a 
vast class of monotonic logics and show that the AGM theory is 
actually applicable to a certain class of logics much wider than the 
original class studied by AGM. Our study will also uncover 
interesting theoretical properties of the AGM theory, as well as its 
connection with other formalisms, like the foundational model and 
lattice theory.
In the fourth part of this dissertation, our focus will be further 
narrowed so as to refer to the main ontology representation 
formalisms, namely DLs and OWL. We will provide specialized 
conditions under which members of these families of logics are 
compatible (or incompatible) with the AGM theory, as well as several 
heuristics and results allowing one to determine this compatibility 
(or incompatibility) for logics not covered by this work. The main 
results of this part of our study is that certain DLs, as well as 
OWL, are not compatible with the AGM theory, while in other DLs the 
AGM theory can be successfully applied for the development of 
ontology evolution algorithms; for the latter category of DLs, an 
algorithm (operator) performing contraction that is compatible with 
the AGM theory is described.

Supervisor:
Dimitris Plexousakis,
Associate Professor of Computer Science
University of Crete