I saw on the Simile mailing list recently a thread discussing possible uses of relationships between user-supplied tags in an information system. This idea is intriguing to me. I've long believed we don't use the relationships recorded in our library-land controlled vocabularies in our systems for end-users to anywhere near their potential. A digital library collection I've been involved with demonstrates ways in which we might use these relationships. The methodology used is documented in this paper.
Yet I'd never thought about relationships for folksonomic vocabularies before. I think it's a fantastic idea, however. The same strategies for improving end-user discovery based on term relationships can be used no matter where these relationships come from. Relationships determined by methods such as this could be used in the same way human-generated relationships in a formal thesaurus could be used. I wonder if these relationships might be even more important in a folksonomic environment, as a method by which the vocabulary control us library folk hold so dear could be achieved.