Querying with Term Similarity

OntoLib provides multiple routines for computing the similarity of terms, given an ontology. For the “classic” similarity measures, the Similarity interface provides the corresponding interface definition.

Here is how to compute the Jaccard similarity between two sets of terms.

// final HpoOntology hpo = ...
JaccardSimilarity<HpoTerm, HpoTermRelation> similarity =
    new JaccardSimilarity<>(hpo);
// List<TermId> queryTerms = ...
// List<TermId> dbTerms = ...
double score = similarity.computeScore(queryTerms, dbTerms);

The Resnik similarity is a bit more complicated as it requires the precomputation of the information content.

// final ArrayList<HpoDiseaseAnnotation> diseaseAnnotations = ...
InformationContentComputation<HpoTerm, HpoTermRelation> computation =
    new InformationContentComputation<>(hpo);
Map<TermId, Collection<String>> termLabels =
    TermAnnotations.constructTermAnnotationToLabelsMap(hpo, diseaseAnnotations);
Map<TermId, Double> informationContent =
    computation.computeInformationContent(termLabels);
PairwiseResnikSimilarity<VegetableTerm, VegetableTermRelation> pairwise =
    new PairwiseResnikSimilarity<>(hpo, informationContent);
ResnikSimilarity<HpoTerm, HpoTermRelation> similarity =
    new ResnikSimilarity<>(pairwise, /* symmetric = */true);
// List<TermId> queryTerms = ...
// List<TermId> dbTerms = ...
double score = similarity.computeScore(queryTerms, dbTerms);