A set theoretical frame based on Svenonius's theory of bibliographic entities is the departure point for this short talk on entity description. This talk will briefly show how properties of bibliographic entity descriptions may be identified using a frequent pattern data mining algorithm over targeted sets of existing metadata descriptions. The MARC21 corpus used in this case was comprised of clustered sets of publishers and publisher locations from the library MARC21 records found in the Platform for Open Data (POD). POD is a data aggregation project involving member institutions of the IvyPlus Library Confederation and contains seventy million MARC21 records, forty million of which are unique.