1984 Reports

Understanding Hierarchically Structured Objects

Wasserman, Kenneth

Natural and artificial hierarchical systems are pervasive. There is a strong need on the part of researcher8 in artificial intelligence and other areas of cognitive science to have mechanisms for "understanding" such systems. Furthermore, computer programs often represent data in a hierarchical form and they would greatly benefit from a technique that would allow them to dynamically build such classification systems from this data. This paper presents a formalism for describing hierarchies and then uses this formalism to explore the issue of inheritance from multiple sources which is of particular importance in hierarchy understanding.