I-SSA: Interaction-Situated Semantic Alignment

Manuel Atencia

Artificial Intelligence Research Institute, Barcelona, Spain.

07/12/2009 à 14h00

Amphithéatre Jean Kuntzmann, Campus universitaire, Saint-Martin d'Hères


The Semantic Web was envisioned as an extension of the Web in which information is given well-defined meaning, better enabling computers and people to work in cooperation. According to this vision, ontologies are commonly taken as a prerequisite for successful interaction. By adopting this stance, meaningful communication between separately engineered software agents relies on a priori commitments to a shared conceptualisation of the application domain, which explicitly specifies what communicated terms shall "mean". Unfortunately, it is often impossible to reach global semantic agreements because precise explicit specifications of meaning are very costly. In addition, according to modern hermeneutics meaning is ultimately interaction-dependent and relative to the background brought into the communication by interacting agents, which cannot be fully de-contextualised. Ontology matching may be helpful as a means of escaping from the rigidity brought about by ontologies, but current state-of-the-art techniques, even when they are dynamically performed at run-time, compute semantic alignments in an interaction-independent fashion. Semantic correspondences are established with the aid of external sources such as WordNet, where semantic relations such as synonymy, among others, were determined prior to interaction and independently from it. We argue, hence, for an approach that takes interaction as ontologically more fundamental than meaning. Interaction model protocols provide a set of conventions that regulate the usage of terms in the utterances of an agent community in a similar way as it occurs in simple Wittgenstein language games. In this talk, we will present "Interaction-Situated Semantic Alignment" (or I-SSA), which only relies on shared interaction state transition to resolve terminological mismatches, avoiding thus dependency on a priori semantic agreements. We will provide a mathematical model and show empirical evidence of the effectiveness of this approach.

© | ? | *


Feel free to comment to Jerome:Euzenat#inria:fr, $Id: 2009atencia.html,v 1.3 2017/01/13 19:38:55 euzenat Exp $