An Application of AI Techniques to Structuring Objects into an Optimal Conceptual Hierarchy
Ryszard S. Michalski, Robert E. Stepp · 1981
A method of "learning from observation" is presented which structures a collection of objects into hierarchies of subcategories, such that each subcategory ia characterized by a conJunctive description involving relations on selected object attributes. The conjunctive descriptions sprouting from each node are mutually disjoint and optimal as a group according to a flexibly defined criterion. Each level of the hierarchy is determined by an iterative process which repeti.tively applies a vrsion of the A* search algorithm. Experiments with the program CLUSTER/PAF implementing the method indicate that the obtained hierarchies represent solutions which have a simple conceptual interpretation and which seem to agree well with the way people structure objects.