An Application of AI Techniques to Structuring Objects into an Optimal Conceptual Hierarchy

A method of " l ea rn ing from observat ion" la presented which s t ruc tu res a c o l l e c t i o n of ob jecta i n t o h ie ra rch ies of subcategor ies, such tha t each subcategory la character ised by a con junct ive d e s c r i p t i o n i nvo l v i ng r e l a t i o n s on selected ob ject a t t r i b u t e s . The conjunct ive descr ip t ions sprout ing from each node are mutual ly d i s j o i n t and opt imal aa a group according to a f l e x i b l y def ined c r i t e r i o n . Each l e v e l of the h ierarchy Is determined by an i t e r a t i v e procaaa which r e p e t i t i v e l y app l ies a vera lon of the A* search a lgo r i t hm. Experiments w i t h the program CLUSTER/PAF implementing the method ind i ca te tha t the obtained h ie ra rch ies represent so lu t ions which have a simple conceptual I n t e r p r e t a t i o n and which seem to agree we l l w i th the way people a t ruc tu re ob jec ta .