On Detecting High-Level Changes in RDF/S KBs

Abstract

An increasing number of scientific communities rely on Semantic Web ontologies to share and interpret data within and across research domains. These common knowledge representation resources are usually developed and maintained manually and essentially co-evolve along with experimental evidence produced by scientists worldwide. Detecting automatically the differences between (two) versions of the same ontology in order to store or visualize their deltas is a challenging task for e-science. In this paper, we focus on languages allowing the formulation of concise and intuitive deltas, which are expressive enough to describe unambiguously any possible change and that can be effectively and efficiently detected. We propose a specific language that provably exhibits those characteristics and provide a change detection algorithm which is sound and complete with respect to the proposed language. Finally, we provide a promising experimental evaluation of our framework using real ontologies from the cultural and bioinformatics domains.

Layered change log model: bridging between ontology change representation and pattern mining

“…Higher-level change representation (in a form of change patterns) is more concise, intuitive and closer to the intentions of the ontology editors and captures the semantics of the change [9]. These ontology change patterns can either be explicitly defined by a user as a higher-level change operators or implicitly given in atomic change log that can be extracted using data mining techniques.…”

“…This is due to the possibility that a subset of a change pattern satisfy the conditions (to be identified) of another category of change pattern [11]. Apart from being more concise, the ontology changes recorded at a higher-level are more intuitive [9]. Identifying the composite changes from the atomic change log give ontology engineer indication about the intent of the applied changes.…”

Semantic Recognition of Ontology Refactoring

“…Highlevel changes of RDF-graphs and version differences (RDF triples) are represented and detected in [5]. They categorize elementary changes like add and delete operations to high-level changes which are similar to refactoring patterns.…”

“…The need to detect high-level changes is already stated in [1,4,5]. High-level understanding of changes provides a foundation for further engineering support like visualization of changes and extended pinpointing focusing on entailments of refactorings rather than individual axiom changes.…”

“…There are different approaches that detect ontology changes by a syntactical comparison like in [4,6] or the combination of adding and deleting RDF-triples to high-level changes in [5]. A structural comparison using matching algorithms is considered in [7].…”