World Report

Report on the Levels of Reality workshop: Bolzano (Italy)

26-28 September 2007

By Claudio Gnoli and Enzo Cesanelli

Ontologies, the knowledge organization systems now widely used in knowledge management applications, take their name from a branch of philosophy. Philosophical ontology deals with the kinds and the properties of what exists, and with how they can be described by categories like entity, attribute, or process. Readers familiar with facet analysis will notice some analogy with the "fundamental categories" of faceted classifications, and this resemblance is not accidental. Indeed, knowledge organization systems use conceptual structures that can be variously reconnected with the categories of ontology. Though having more practical purposes, the ontologies and classifications of information science can benefit of those of philosophy (see Poli in Proc. ISKO conf. Washington 1996).

Ontology can be subdivided into three sub-branches: descriptive ontology, devoted to collecting the data which comes from reality; formal ontology, devoted to filtering, codifying, and organizing those data according to categories; and formalized ontology, translating these organizations in terms of formal logic, with axioms and deductions. The last approach is widely used in the application of ontologies to computer science, leading to build knowledge bases that can be used in artificial intelligence. One general formalized ontology is developed in the DOLCE project, led by Nicola Guarino (National Research Council, Italy), editor-in-chief of the "Applied ontology" journal, who also attended the seminar.

However, John Sowa (Vivomind, USA) argued in his speech that the formalized approach, already undertook by the pioneering Cyc project now having run for 23 years, is not the best way to analyze complex systems. People don't really use axioms in their cognitive processes (even mathematicians first get an idea intuitively, then work on axioms and proofs only at the moment of writing papers). To map between different ontologies, the Vivomind Analogy Engine relies less on axioms than on finding analogies in their structures. Analogy is a pragmatic human faculty using a combination of the three logical procedures of deduction, induction, and abduction [presentation]. Guarino comments that people can communicate without need of axioms as they share a common context, but in order to teach computers how to operate, the requirements are different: he would not trust an airport control system working by analogy. Sowa answers that the critical requirement would be extensive testing.

Information applications are also addressed by Heinrich Herre (University of Leipzig, Germany) with his group working on GFO: General Formal Ontology, a development of the medicine ontology OntoMed. Modeling of biomedical domains employs the notion of levels of reality, and requires three kinds of categories: for the principal object of a domain, for its taxonomies, and for its aspects or facets [presentation]. Again, this shows resemblances with bibliographic classification. It is another signal that philosophical ontology should be taken into account in knowledge organization, and that much remains to be done for a fruitful integration of its experience with those of other fields like computer science, information architecture, psychology, linguistics, library and information science.

This is an extract of the full report, which can be read at Occam's razor and Poli's beard.