A new approach to the construction of general persistent polyhierarchical classifications is proposed. It is based on implicit description of category polyhierarchy by a generating polyhierarchy of classification criteria. Similarly to existing approaches, the classification categories are defined by logical functions encoded by attributive expressions. However, the generating hierarchy explicitly predefines domains of criteria applicability, and the semantics of relations between categories is invariant to changes in the universe composition, extending variety of criteria, and increasing their cardinalities. The generating polyhierarchy is an independent, compact, portable, and re-usable information structure serving as a template classification. It can be associated with one or more particular sets of objects, included in more general classifications as a standard component, or used as a prototype for more comprehensive classifications. The approach dramatically simplifies development and unplanned modifications of persistent hierarchical classifications compared with tree, DAG, and faceted schemes. It can be efficiently implemented in common DBMS, while considerably reducing amount of computer resources required for storage, maintenance, and use of complex polyhierarchies.