gov.llnl.ontology.clustering
Interface ClusterSimilarity

All Known Implementing Classes:

LocalitySensitiveWordSpaceClustering, WordSpaceClusterSimilarity

public interface ClusterSimilarity

This interface provides a mechanism for determining the similarity between two words after their semantic representations have been clustered. Similarity between two words is based on two factors: 1) the semantic representations and 2) cluster assignments. Two terms that are assigned to diffefent clusters should have a lower similarity, possibly 0, than two terms assigned to the same cluster.

Implementations should maintain a word to vector mapping, based on a SemanticSpace algorithm, so that semantic vectors are easily retrievable.

Implementations should implement toString so that other codes can query the label of the clustering method used.

Implementations should also implement Serializable so that other codes can store and restore instances of this interface.

Author:

Keith Stevens

Method Summary

double

getTermSimilarity(String term1, String term2) Returns the similarity between term1 and term2.

Method Detail

getTermSimilarity

double getTermSimilarity(String term1,
                         String term2)

Returns the similarity between term1 and term2. This similarity should be based on part on the clustering assignments.

A common approach is to limit similarity to be between terms in the same cluster. All other term pairings have a similarity of 0.