Similarity Methods

This section documents the theoretical background of the similarity methods offered by bioterms-cli similarity calculate (see Constructing Database), how each score is derived mathematically, and how the calculation is actually carried out against the concept graph. For a description of the API surface that serves these scores once calculated, see Similarity instead.

Every method calculates a pairwise score between two concepts belonging to the same target vocabulary, using the annotation relationships between that target vocabulary and one corpus vocabulary. A score is only ever meaningful between two concepts of the target vocabulary; the corpus is used purely as the evidence base from which each method derives its statistics, and does not itself need to be similarity-scored.

• Relevance Method
• Co-Annotation Vector Method
• Weighed Relevance Method