FOIS conference report

Second day of the conference in Formal Ontologies in Information Systems held in Torino. Today was completely into the ontology theory and the more academic approaches to the matter. After a beautiful presentation by Amie Thomasson on the Methods of Categorization, we moved on the first session on the definition of the Categories. The following session was on the perspectives of using ontologies followed on some methodological notes. The full report in the extended.

My general impression is that the ontology theory needs some more practical examples; it needs to reflect more on the common sense realism; and also it needs to integrate more some cognitive semantics approach. Generally there is the need for examples, ontologies concrete and applied to specific and detailed fields so to constitute case studies. So far there has been to much attention to trying to find universals and there is a continuos debate on what exactly is an universal and how to categorise things.

TITLE OF PAPER: FOIS Formal Ontologies in Information Systems

DATE: November, 4, 2004
LOCATION: Torino Incontra, Centro conferenze, Torino, Italy


Welcome addresses
Nicola Guarino. We started this conference 6 years ago. Nowadays ontologies is very well known. This word is often used to discuss about semantic interoperability in information systems.

This discipline is looking at representing in a rigorous and universal ways content. Fois is interdisciplinary. Achille Varzi is a phylosopist.

Magnifico Rettore of University of Turin. This field is very interdisciplinary. The growing interest on the semantic web is a proof of the importance of ontologies in informatics systems.

Invited Specker: Amie L. Thomasson (University of Miami): Methods of Categorization

Philosophy. Reasons to want a Category Systems: helping developing an ontology systematically. Determining ontological nature of entities of specific kinds. Uses in information science and biomedical sciences.

Past embarrassments. Lack of consensus on what the basic are. Lack of consensus on the form a system of categories should take. Failure to determine the appropriate methods and criteria.

Past Category Systems
-Aristotle: Categories of real entities, not though or language; First category: substance, corresponding to grammatical subjects; further categories drawn out via language: by what could be said of or in a substance; forms of suitable questions and answers. All categories are same-level highest; no single subsuming category. Worry: why assume categories of language can reveal categories of real entities in the world?
-Kant: uses categories of thought (judgment) to distinguish categories. Guaranteed correlation between categories of understanding and categories of objects of possible knowledge, since the understanding imposes categories on objects known. Categories are of mere phenomenal objects.
-Some Contemporary Category Systems: Ingvar Johansson: Seeks a realist theory of categories regarded as real aspects of being; method: unmediated by thought or language; successive abstraction. Worries that remains: does successive abstraction presuppose categories to identify the individuals from whom we abstract? Could we abstract further to get a topmost category, e.g., entity? Other options that do so: Chisholm, Hoffman & Rosenkrantz.

Outcome of Survey: Category Fatigue. What is the right: set of categories; form of a system of categories; fundamental division (if any)? method for distinguish?

Criteria of adequacy: categories should help to answer: What exists? What sort of thing is this? Minimal criteria of adequacy: categories must be exhaustive (they should not force dichotomy); categories must be mutually exclusive;

Methods of categorization:
-A. The feature-negation method: For any feature F. Yields two highest level categories. Further subdivisions regarding other features may not be possible. May apply the method across … Advantages: principle of non-contradiction ensures mutual exclusivity. Law of excluded middle ensures joint exclusivity. Difficulties: which features to choose? Can being not-F a fundamental category? Do such categorizations lead to widespread category mistakes?
-B. The method of Absurdity Detection: Husserl: two terms are of different (grammatical) meaning-categories if substitution turns sense into nonsense. Ryle: two expressions are of different (conceptual) categories if substitution turns sense into absurdity. Difficulties: J.J.C. Smart: Ryle’s criterion can be used to show that there is a category difference between any two expressions.
-C. Methods of distinguishing identity and existence conditions: Frege: names attempt to designate individual objects. Husserl’s nonsense test may distinguish different grammatical categories. Fredge/Dummett: Names are those that: -have associated criteria of application and have associated criteria of identity. Criteria of identity: normally supplied by associating the name with a basic sortal term, such that the entity exists only provided the basic sortal term continues to apply to it. Two names entities are identical only if they name things of the same basic sort’s criteria of identity. Ontological categories: categories of objects-the correlates of names. Dummett: two terms and the object they name are of the same category if they are associated with the same criteria of identity (application conditions may vary).

New proposal: consider both conditions of identity and application conditions in formulating categories. Then categories may be … Features: since species and genus sortals apply to the same entity, each species sortal is guaranteed to share the same criteria of identity as its genus sortal.

Consequences of the proposal: different identity conditions -> different category, thus we can categorically distinguish

Answers to methodological question: ontological categories are (like Aristootele’s) drawn out via considering language (grammatical category of names; then associated …)

Exhaustiveness may not be a good criteria.

Session 1: Categories
<> Thomas Bittner Individuals, Universals, Collections -> On the Foundational Relations of Ontology. He tries to define the relations between individual, universals and time. Three categories. Time dependent and independent relations.

<> Fabian Neuhaus, Pierre Grenon and Barry Smith (Saarbrucken) -> A formal theory of substances, qualities, and universals. Aristotelian Realism: assumption A there is a reality which is independent from language our thinking our culture. Assumption B. If we understand the general structures of reality … Universals (the pattern of reality, universals are the ontological counterparts of the general terms of our taxonomy, universals are instantiated by particulars) vs. particulars

<> Jonathan Simon (NYU Dept. of Pylospophy)-> How to be a bi-categorialist
Bicategorialism (Dualism): perdurants persists by having different parts at different times. Endurants persists by being located in full at multiple times.
[mereological relations]

Session 2: Perspectives
<> Barry Smith -> Beyond Concepts: Ontology as Reality Representation
UML is full of mistakes: the linguistic reading yields to good readings but it fails miserably when it comes to relations of other types. The engineering reading of meaning. The ontology reading: concepts are nor creatures of cognition or of computation. We have to avoid using the term concepts for the term meaning or universals. We have to conceive ontologies instead as representations of reality.

– what about common sense?
– language is used in an undisciplined way, nevertheless is used successfully because it is used in a very rich environment.

<> Simon K Milton (University of Melbourne) -> Top-Level Ontology: The problem with Naturalism.
Does high level ontology really matter? top-levels vs. domains specific. The use of top-levels ontologies ensure coherence. But, do we care in information systems? Common sense realism is much more useful in accommodating real problems.

<> Joseph A. Goguen -> Ontology, Society and Ontotheology.
There is much hype on semantic web but what is realistic? Large databases have lots od problem because data gets confused. Pragmatism says that theories are not universal truth but are useful approximation.
[book Sorting Things out, MIT press 1999]
Real world classifications are complex and heterogeneous, highly political and evolving.
Roch: basic level concepts
— blending, image schemas
We need better theories that are more grounded. Ontology need less hype. We should respect more the authenticity, heterogeneity of beings.

We should look at ontologies that have been used for years, there must be something good in there, so maybe we can learn some lessons from there instead of trying to impose some philosophical theories from above.

The music is not in the script the music is in the experience. I like this phenomenological approach and that therefore it is a metonymie on using ontologies looking more at the real usage of these tools. We need to distinguish between ontologie and phenomenology.

Leave a Reply