Definition: SimilarityMeasure

The concept of a SimilarityMeasure in the PhaseLibs describes an interface to measure a distance (of any kind) between two ontology entities (Frames?). These entities can be taken from the same, or from two different ontologies. The resulting value is rquired to be in the interval [0..1].

A SimilarityMeasure can use any information available (i.e. accessible via the ontology model interface) about the given entities (Frames?). Furthermore a SimilarityMeasure can calculate a matrix of similarities to enable the algorithm to work more efficiently on large batches.

This basic concept is represented by the interface de.dfki.km.phaselib.model.evidence.SimilarityMeasure.

Extended Concepts

In order to allow for more specific information to be gathered from the algorithms, several extensions of this SimilarityMeasure concept exist.

asymmetric SimilarityMeasure

There is the concept of the asymmetric SimilarityMeasure, which allows the values of similarity(entityA, entityB) and similartiy(entityB, entityA) to be different. It is represented by the interface de.dfki.km.phaselib.model.evidence.AsymmetricSimilarityMeasure. This interface extends the SimilarityMeasure by methods to query this directed similarity, leaving the old methods to some sort of joint similrity, to be calculated from the two directed ones. A common approach to this calculations would be the minimum of the two similarities.

extended SimilarityMeasure

Furthermore, there are algorithms (like SimilarityFlooding) that additionally provide information about a dissimilarity of entities. To allow for such information, the interface de.dfki.km.phaselib.model.evidence.ExtendedSimilarityMeasure requires methods for dissimilarity querying.