Some remarks on the possibility of extending the theory of polyrepresentation to the study of the relevance and qualities of 'things' represented in information collections