wiki:similarity_GraphMatching

Version 5 (modified by spillman, 18 years ago) (diff)

--

TracNav?

SimilarityMeasure: Graphmatching

Developer: Barbara Spillmann

Description

The Graph Matching module can be used to determine similarities of ontologies with respect to the structures of the ontologies under consideration. An ontology is mapped to a graph where each ontology element (concepts, relations, attributes) is represented as a vertex in the graph. Once such graph structures have been extracted, maximum common subgraphs (mcs) can be determined. An mcs refers to structurally equal parts of two graphs, or finally of two ontologies.

Originally, the definition of the maximum common subgraph of two graphs is strict in the sense that only exact subgraphs regarded. Applied to ontologies this is really not uesful as exact subparts of ontologies are rare. Therefore, a weaker definition of the mcs is used: subgraphs are defined to be common if they are similar enough. The similarity of subgraphs can be measured by various measures and possibly controlled by a threshold parameter.

Characteristics

The calculation of the maximum common subgraph is very time-consuming.

Evaluation/Performance

Some evaluation tests on the Benchmarks have been performed. You can inspect the results here.

Specification

Initialisation

The SimiliarityMeasure main class is

de.dfki.km.phaselib.TODO

Parameters

TODO

Dependencies

TODO

License Issues

TODO