Relatore : Marie-jeanne LESOT

Co-relazione : LAUDY Claire (Thales)

Representation, simulation and exploitation of knowledge in the conceptual graphs formalism

The work carried out in this thesis is situated in the framework of conceptual graphs, which constitutes a formalism for graph-based knowledge representation with advantages of expressiveness and interpretability, as well as formal guarantees and efficient reasoning.
In this context, we propose various theoretical, algorithmic and practical contributions. First, we address the problem of exploiting a conceptual graph base in order to infer new knowledge, or generalize existing ones, and we propose to that aim an algorithm for the extraction of frequent patterns of interest taking into account the specificities of conceptual graphs. Then, we are interested in need of benchmarks for the validation of such algorithms, and we propose simulation algorithms for the generation of conceptual graph bases, from translation and from ontological constraints. Finally, we consider the set of extensions that increase the expressiveness of the conceptual graph formalism, in particular to represent imprecise knowledge, by combining them with fuzzy set theory. In this context, we propose a comparative discussion of fuzzy conceptual graphs state-of-the-art proposals, including two taxonomies that organize the studied models according to the impacted part of the formalism and according to their common points and differences with the existing cases, adopting a contrastive approach.

Pubblicazioni 2021-2022