Article Preview
Top1. Introduction
Domain knowledge capture has always been of crucial interest for humans and computer systems in several fields. Semantic formalisms such as ontologies allow holding this knowledge as well as sharing and understanding data. They also guarantee data interoperability in heterogeneous environments. However, knowledge is in continuous change in line with the evolving real-life requirements. Thus, an ontology has to follow such evolution and change accordingly. This process, known as ontology evolution is “…the process of managing ontology changes by preserving the consistency of the ontology with respect to a given notion of consistency…” (Haase, 2006, p. 126). This definition reveals that ontology consistency management is an important task in the ontology evolution process (Khattak, Batool, Perve, & Mehmood Khan, 2013; Mahajan & Kaur, 2015). Ontology is commonly called consistent if it is free of logical contradictions (Flouris, Huang, Pan, Plexousakis, & Wache, 2006; Kalyanpur, Parsia, Horridge, & Sirin, 2007). It is equally called inconsistent if it violates the syntactic constraints of the language (Stojanovic, 2004) or knowledge modelling guidelines (Haase & Stojanovic, 2005; Poveda-Villalón, Suárez-Figueroa, & Gómez-Pérez, 2012).
To address the different types of inconsistencies, several approaches and tools have so far been designed (Zablith et al., 2015). Nevertheless, most of them cope with inconsistency at a late stage of the ontology engineering process. Indeed, inconsistency is checked and repaired in an a posteriori way. This is carried out either after the effective application of changes (Djedidi & Aufaure, 2010; Flouris et al., 2006; Haase & Stojanovic, 2005; Kalyanpur et al., 2007) or relying on already adapted, stored ontology versions (Copeland, 2016; Poveda-Villalón et al., 2012). According to such a posteriori consistency checking approaches, inconsistency has already propagated to dependent artifacts (i.e., parts of ontology, related ontologies and applications). Hence, the modeled knowledge becomes meaningless until an ontology repair action is undertaken. At this stage, many rolling backs may have to be conducted so as to restore the ontology to its last consistent state.
This paper addresses these problems by providing a broad view of consistent OWL 2 DL (Web Ontology Language Description Logic) ontology evolution, where the inconsistencies that can arise due to changes on ontology are not only restricted to logical inconsistencies, but also include syntactical invalidities or style issues. An a priori consistency checking approach is adopted. To this end, change kits are defined to anticipate different inconsistency types at an earlier stage. These formalisms are applied to a consistent OWL 2 DL ontology and anticipate potential inconsistencies due to a requested change in order to prevent their occurrence. Hence, change kits provide a means for implementing the change while preserving the ontology consistency. As a consequence, they minimize expert interventions or any rolling back for checking and restoring consistency. A prototype tool is also implemented to support knowledge engineers in preserving a consistent state of an ontology upon each attempted change. It inspects the ontology changes before they are applied, issues appropriate feedback and possibly suggests alternatives.
The remaining of this paper is structured as follows. Section 2 exhibits the state of the art related to the ontology evolution approaches and tools. Section 3 describes the proposed Inconsistency Anticipation Approach. Section 4 presents the developed tool and Section 5 shows an illustrative example of its use. Section 6 discusses the conducted experiments before concluding this work.