Nicola Guarino - Concepts, attributes, and binary relations: representational and ontological issues

Due to their cognitive relevance, frames play an important role in knowledge representation: modern representation languages, such as description logics, offer specific con-structs to describe the internal structure of frames - or, as we may call them, structured concepts. These constructs allow to describe a concept as a “pattern of relationships”. But what is the ontological nature of such relationships? How can we decide whether a single relationship belongs to a frame, playing the role of a structuring relationship? Or, in other terms, what are the “natural boundaries” of a frame? Is there a “grammar” of relations that can or cannot constitute a frame? In the Artificial Intelligence literature, this problem was raised by Bill Woods in a seminal paper, where he considered the following “frame” (in the computational sense):

JOHN
    HEIGHT: 6 FEET
    HIT: MARY

In the intended interpretation of this structure, both HEIGHT and HIT are “links” (i.e., binary relations) that contribute to characterize the JOHN node. Woods’ point was that there is an obvious semantic difference between the two: the former can be considered as an attribute of John, the second cannot, it is just an arbitrary relation. Therefore, the two constructs should be represented in a different way, since, in a sense, HIT does not belong to the JOHN frame.
Despite Woods’ suggestions, modern representation languages typically ignore this issue: description logics expressions rely on primitives called roles, whose semantics is that of arbitrary binary relations. They fail therefore to capture, by means of struc-tured descriptions, the crucial cognitive aspect of frames, that is the fact that they should form coherent conceptual unities, not just arbitrary bundles of relationships. I wrote a paper on this issue more than 15 years ago, whose results have been used in the project which inspired this conference. Not much happened since then in the knowledge representation community, however, except the fact that now a whole interdisciplinary area of research emerged, Formal Ontology, that can help understand-ing and formalizing this problem to a good extent. In this talk I will very briefly introduce the role of formal ontology in knowledge representation, and I will exploit its basic tools for disentangling and understanding the problem above, trying to clarify the subtle ontological and linguistic distinctions and relationships concerning concepts, roles, attributes, and binary relations.

References

N. Guarino, Concepts, attributes, and arbitrary relations: some linguistic and ontological criteria for structuring knowledge bases. Data and Knowledge Engineering 8 (1992): 249-261

W. A. Woods, What’s in a link; foundations for semantic networks. In D. G. Bobrow and A.M. Collins, eds, Representation and Understanding: Studies in Cognitive Science. Academic Press, New York, 1975. Reprinted in R. Brachman and H. Levesque, eds., Readings in Knowledge Representation. Morgan Kaufmann, Los Altos, CA, 1985.