Instance Migration Between Ontologies Having Structural Differences

Abstract

Ontology instance migration is one of the complex and not fully solved problems in knowledge management. A solution is required when the ontology schema evolves in the life cycle and the assertions have to be transferred to the newer version. The problem may become more complex in distributed settings when, for example, several autonomous software entities use and exchange partial assertional knowledge in a domain that is formalized by different though semantically overlapping descriptive theories. Such an exchange is essentially the migration of the assertional part of an ontology to other ontologies belonging to or used by different entities. The paper presents our method and tool for migrating instances between the ontologies that have structurally different but semantically overlapping schemas. The approach is based on the use of the manually coded transformation rules describing the changes between the input and the output ontologies. The tool is implemented as a plug-in for the ProjectNavigator prototype software framework. The article also reports the results of our three evaluation experiments. In these experiments we evaluated the degree of complexity in the structural changes to which our approach remains valid. We also chose the ontology sets in one of the experiments to make the results comparable with the ontology alignment software. Finally we checked how well our approach scales with the increase of the quantity of the migrated ontology instances to the numbers that are characteristic to industrial ontologies. In our opinion the evaluation results are satisfactory and suggest some directions for the future work.

Agent-Based Implementation for the Discovery of Structural Difference in OWL-DL Ontologies

“…In the presented agent-based implementation the process embodies the strategy of automated meaning negotiation proposed by Ermolayev et al in [4] for efficient information retrieval in open and distributed environments. The presented work is a part of the research activity aiming at development and implementation of an efficient methodology of ontology alignment and instance migration in decentralized settings (see also [5] for details on instance migration part of the methodology). The methodology (depicted on figure 1) assumes (semi-) automated iterative process of ontology alignment and instance migration with possible human intervention for checking the correctness and setting up the process.…”

“…Also, this format is the de facto standard for representing alignments for participants of OAEI 2 and the presentation of results in this format allows us to estimate the system using OAEI benchmarks and makes the results comparable to those provided by OAEI. The second type of output is the representation of the alignment in the form of structural difference and generation of respective transformation rules for the next step -the migration of ontology instances (the problem statement of ontology instance migration and the description of solution based on transformation patterns and rules are described in detail in [5]). At the second step agents use Instance Migration Engine in order to transfer instances between ontologies and based on the transformation rules generated at the first step.…”

“…At the second step agents use Instance Migration Engine in order to transfer instances between ontologies and based on the transformation rules generated at the first step. Problem cases occurring during the migration are recorded into the migration log (possible problems and collisions during the migration are considered and analyzed in [11] and [5]). Finally, the target ontology agent asks a reasoner for the consistency check of the obtained ontology .…”

“…As noted above, on the basis of the obtained alignment an agent forms the set of transformation rules to convert and transfer the ontology instances. More information about the transformation pattern and rule language can be found in [5]. For consistency checking in our implementation we use the Pellet reasoner 6 .…”

“…According to the two-step character of the alignment process we can measure the quality after each step on separate iterations. For these purposes we plan to use Precision, Recall and Fmeasure metrics (the description of these metrics for measuring the quality and completeness of ontology instance migration is given in [5]; the use of the metrics for assess ontology alignments is quite similar). We plan to use OAEI ontologies and datasets as a test-bed for our experiments.…”