The semiotic approach to information exchange is fully included in our research project. However, lack of time and resources does not allow one to work enough on this. It is thus rather a prospective activity so far.
How ensuring the intelligibility of the content of a message between two users? It is possible to control the correct apprehension of the message by a computer (because we know that the computer treatment is correct and complete with regard to the semantics of used language). It is very difficult to know which interpretation a user will favour. Several convincing instances of that problem have been produced (in particular, concerning electronic peer-review software which do not allow the receiver to interpret correctly the readers' comments) [Euzenat2000d]. The meaning attributed to an assertion by a human user is generally a restriction of the one given by formal semantics [Euzenat2003c]. This can be refined by saying that the user will select a subset of the models of the initial representation. The example of different knowledge representation formalism is very important in collaborative ontology construction and the semantic web.
The traditional approach to that problem is to ask for more axioms from the part of users, restricting the set of models. Work on consensual ontologies [Guarino 1998] aims at solving this part of this problem through "Ontological commitment" which consists exactly of adding the axioms of the ontology. However, this is neither always possible nor easy for the users. An alternative approach consists of taking into account the rules that govern the model selection process. These rules can be semiotic (because some constructions have a particular interpretation), rhetoric (because the message is built so that a particular interpretation is forced from the receiver point of view) or pragmatic (because the interpretation is tied to the enunciation context).
We are looking for a formal approach of these disciplines (or part of them) that could be informatically described. Others have found this problem in building an editor for the semantic web [Bechhofer 2001]. In a general way, the considered problems involve semantics of knowledge representation, semiotics and natural language semantics (because our goal is the preservation of meaning through the information supply chain). However, the state of the art in semiotics leaves little hope in a short-term solution. There exists however, some basis exploiting in a reasoned manner structuralists methods [Andersen 1990]. The more immediate link is, again, with "rhetorical structure theory" used in text generation [Mann 1988] because it offers a computerised treatment of the structure behind a formally or informally defined text.
A possible way would consist of adding, together with syntax and semantics, the sign interpretation rules used by the sender and receiver so that the message is presented to the receiver in a way in which (s)he will correctly interpret the message. These techniques are alike those of user profiles (or device profiles in CC/PP) applied to language interpretation instead of preferences (or capabilities). For the time being, we are mainly looking for partners in human and social sciences for studying this. For that purpose, we are building examples showing the relevance of the approach.
For instance, one can express knowledge as a class hierarchy and translate it to an interoperability language for communicating it to someone else. If that language expresses the knowledge under the form of clauses (while preserving the interpretation of the assertions), the sender might have difficulties to recognise this semantically equivalent result. Thus, when knowledge is translated between several representation systems, good understanding cannot be taken for granted. The main explanation is that human understanding depends on the form although correctness of computer manipulation only relies on semantics. Hence, a semiotic (or pragmatic, as said in linguistics) treatment must be attempted for easing the reconstruction of meaning. This treatment comes as a complement to the sheer semantic treatment presented above.
A more elaborate track, called algebraic semiotics, has been opened by Joseph Goguen [Goguen 1999]. In the continuation of his work on institutions, it consists of considering the representations in relation with their use. Algebraic semiotics integrates a (sign) system, its syntax, its semantics and, as far as possible, the rules of interpretation that a user will use in order to interpret the representation. A morphism transforms such a system (the source) into another (the target) supposed to be more suited to a particular task. Algebraic semiotics studies the sign systems and their transformations in the framework of category theory. This would provide a unified standpoint for considering the manipulation of information.
Goal: Developing an understanding of sign system interpretation that could be represented in and processed by a computer in order to ensure a semiotic-level interoperability. Illustrating this on a few precise examples.