This would be an alternative to informationContent for calculating semantic distance between two terms
Note that OLS now has vector embeddings, I think most of UG is in OLS, so could harvest from there, also use the same cache strategy as James
cthoyt/ols-client#9