Efficient Retrieval from Hierarchies of Objects using Lattice Operations

Managing large numbers of complex objects or descriptions requires sophisticated storage and retrieval mechanisms. Hierarchical classification techniques have proven to be useful for managing complex objects in practice. Previous classification methods have not used all the information recorded in hierarchies. We show how lattice operations can be used to navigate around a hierarchy. This is achieved by plunging the hierarchy into a boolean lattice of binary words. Greatest lower bound (GLB) and relative complementation (BUTNOT) lattice operations on the hierarchical structure (rather than the objects or descriptions themselves) are used to focus the search. Experiments show the number of objects compared when classifying objects using these techniques is significantly reduced.