1. THEORY OF MEANING IN ONIONS

1.1 A semiologic theory of conceptualization

1.1.1 In praesentia vs. in absentia signs At the core of semiologic theory is the distinction between praesentia and absentia, provided by Saussure [1970]. Given a meaningful linguistic chunk -- a word, a phrase, a morpheme, a sentence -- which he called sign, such a sign could be in praesentia if it is considered as part of an actual text we are dealing with, while it could be in absentia if it is a relevant part of possible texts that can be associated to another in praesentia sign.

In logical terms, given a domain of signs D and a domain of texts T, constructed within D, for

The main puzzle of this distinction for logical semantics is the intended meaning of the predicate RI. That is, RI is meant as the possible connections that a sign can stir up in an ideal knowledge agent in a given culture and in a given domain: in other words, tj is any text which contains a relevant interpretation for that sign, given the pragmatic rules of a linguistic community.

The underlying intuition of these definitions is that a conceptualization of a domain is a textual interplay of signs. This also implies that whatever ontology will grasp some interpretations of a conceptualization (but the question if they are valid, or relevant, is another story, cf. par. 2.1).

1.1.2 Semantic fields A conceptualization focalizing on a given sign, or on a finite set of signs, is represented by a subset of signs: such a subset individuates a semantic field. Hence, a semantic field-- on its turn -- individuates the set SF of in praesentia or in absentia signs, relevant for a given set of in presentia signs: given the set of signs S < ti, it holds that:

A kind of restricted semantic field is the lexical field: this one includes all the in absentia signs from another specified set of texts (not all the possible ones); thus the lexical field (LF) version from the previous formula states that y must be part of a tj that is included within a soundly selected corpus T* of texts from a domain:

This definition evidentiates the dependence of a lexical field from the local cultural context (see next paragraphs), represented by T*, that (for instance) is constituted by medical dictionaries and medical experts in our experiment on medical domain.

1.1.3 Theories as in absentia semantic fields The textual interplay of signs can be generalized to consider the interplay of theories: what actually happens is that in absentia signs are vehiculated by a theoretical kinship, ie, by their kinship within other semantic fields, already encoded in culture, language, operations carried out by people. Fig.1 metaphorically shows this idea: onion shelves are the conventionalized links from a sign to in absentia theories. In the next paragraphs, it is suggested that such interplay of theories as semantic fields is motivated.

1.1.4 Contextual binding on conceptualization A similar theory to Saussure's was contributed by philosopher Peirce, whose notion of interpretant constitutes the medium between his sign and external world, intended as a dynamic object [Peirce, 1980; Eco, 1984]. An interpretant is any sign or group of signs which is used to interpret another sign and/or a dynamic object.

The interpretant is a notion quite similar to Saussure's in absentia sign; only, it is considered dynamically, as dependent on the human active involvement in real world processes and human attempts to interact with environment by interpreting it through the inferential process of (semiotic) abduction. In other words, Peirce stresses the issue that a context faces a perceiving individual with an interpretive need, which is accomplished through his/her navigation within the ideal knowledge (either theories or operational knowledge, cf. par. 2.3): a dynamic net of interpretants, which, among others, features linguistic signs.

1.1.5 Local vs. global conceptualization In the ideal knowledge, the interplay of interpretants acts as the global, dynamic conceptualization which motivates the mutual comprehension of humans.

In fact, only a local context can motivate a given ontology which aims at schematizing a conceptualization.

A remark has to be made on what a context is meant to be. In fact, a spatial region or a situation including a perceiving individual is a context in an external sense, while a dynamic region in the knowledge net of a cognitive subject is a context in an internal sense. Finally, a domain of the ideal knowledge of an intersubjective culture is a cultural context.

1.1.6 Gestaltic bindings But how to interpret a local context inside the global conceptualization, without modelling the last in its entirety? Here comes the need for a grounding of signs in their gestaltic bindings.

There are cognitive and external constraints on global conceptualization and, derivatively, on a context. A source of gestaltic constraints is the actual structure of the world [Petitot&Smith, 1991], at least the structure that directly interacts with humans: the ordinary, or 'common sense' structure [Varzi&Casati, 1994; Poli, 1992], which can be described by theories concerning universals, or invariants, of cognitive perception, such as wholeness and parthood of objects [cf. Simons, 1987], connectedness, strata of reality (material, biologic, psychological, socio-cultural) [cf. Hartmann, 1966].

Another source of gestaltic constraints comes from cognitive schematization [Lakoff, 1990; Langacker, 1991; Talmy, 1995], which intervenes to impose form on perception data, as well as upon syntactic and grammatical structure, and even upon interpretation paradigms. Especially kinaesthetic image schemata, such as up/down, front/back, containment, configuration, path, link, force dynamics, pave the way for perceiving external world, for ordering words and understanding syntactic relations.

In other words, common sense and schemata should be a means for constraining the complex structure of global conceptualization, thus performing a restriction on the structure of a local context as well.

1.1.7 Gestaltic constraints as field operators If gestaltic constraints are common to both global and local (contextual) conceptualization, they could be seen as operators which structure semantic fields. This is the approach we have adopted, making a strong distinction between structural signs and structuring signs (operators) within a semantic field. In par. 3., they will be referred as sorts and relations.

Incidentally, we remark that the signs resulting from the integration made in the medical domain through ONIONS methodology converge with some other frameworks which share similar theories of meaning (for example, the semantic fields and operators types in [Miller&Johnson-Laird, 1976]).

1.2 Individual vs. constructive reference: how a theory of meaning may influence the ontological commitment

A potential source of confusion about formal ontology derives from the relation among linguistic items, predicates and objects.

In logical, non-semiological semantics, a linguistic item is assumed to correspond to a pre-defined, pre-constructed, independent object via an implied concept. This classic assumption is well-suited to the use of an ideal individual as the argument satisfying a predicate, which translates into a referentialist approach to formal logics (Fig. 2). This makes one run the risk of confusing names, concepts, and objects, which are all recalled by the same linguistic item.

Fig.2: The logical semantics referential approach: one is free to identify a linguistic item with a concept or with an individual object by predicating something of an individual variable.

On the contrary, semiology assumes no pre-defined object, because the correspondence is always to a local construction of linguistic items within a semantic field, having an ultimate reference in the ideal knowledge net of a culture, community, etc.

When a formal logic is used within this approach, the individual remains in the background (the "dynamic object"), and predication operates on "sorts" only (Fig.3). This approach prevents us from identifying names or "concepts" with the object: our predications are explicitly bounded within the textual interplay of linguistic items.

Fig. 3: The semiological approach to reference: one cannot identify a linguistic item with anything else, since a sign is predicated of other signs, which only indirectly refer to a dynamic object.

1.3 Conceptualization for ontological engineering

For reasons of clarity within the community of AI or KA people, and since sign has a special medical meaning as well, we decided to use concept instead of sign, but with the warning that its meaning is intended on the basis of the theory of meaning described here.

To sum up briefly this discussion, we claim that ontological engineering could successfully account for a conceptualization only if it can select a semantic field (a lexical field, in the perspective of this work), to partially account for the local contexts underlying a conceptualization, and if it can apply intersubjective theories which describe both the cognitive and external constraints of a semantic field.

At this point, we can define a conceptualization as: the definition of a semantic field, by encoding the contextual and gestaltic bindings of a concept (or of another semantic field), and of its associated theories.

In par. 2. an operative approach to conceptualization and ontology is outlined which encodes a context by means of lexical fields and encodes gestalts by means of general theories.


The Ontology Group Home Page (under construction)


for information about the maintenance of these pages, write to aldo gangemi