= Ontology Matching Using Graph Matching --- Results =

We have performed experiments on the Benchmark tests. The algorithm applied is the the MaxComSubgraphAlgorithm with a VertexTypeEquality measure (two vertices are equal if they are both concepts or relations or attributes) and a Longest Common Subsequence (lcs) similarity measure for the vertices' labels. These two measures are concatenated with a conjunction.


To view the detailed precision and recall values for the Benchmark tests, as well as the Fallout, the F-Measure and the Overall measure, see this [wiki:GraphMatchingResultsTable table].

The following figures show the F-Measure and the Overall Measure for the maximum common subgraph method. The used threshold with values 1.0, 0.9, 0.8 and 0.7 (in this order from left to right) indicates how similar the vertices' labels are. The value 1.0 refers to label equality according to th lcs similarity measure.

== F-Measure ==
[[Image(mxcs00int-fmeasure.jpg, 24%, border:solid 2px gray)]]  
[[Image(mxcs01int-fmeasure.jpg, 24%, border:solid 2px gray)]]
[[Image(mxcs02int-fmeasure.jpg, 24%, border:solid 2px gray)]]
[[Image(mxcs03int-fmeasure.jpg, 24%, border:solid 2px gray)]]


== Overall Measure ==
[[Image(mxcs00int-overall.jpg, 24%, border:solid 2px gray)]]
[[Image(mxcs01int-overall.jpg, 24%, border:solid 2px gray)]]
[[Image(mxcs02int-overall.jpg, 24%, border:solid 2px gray)]]
[[Image(mxcs03int-overall.jpg, 24%, border:solid 2px gray)]]




Back to the [wiki:similarity_GraphMatching Graph Matching] main page.