Conceptual Structures Represented by Conceptual Graphs and Formal Concept Analysis

Conceptual Graphs and Formal Concept Analysis have in common basic concerns: the focus on conceptual structures, the use of diagrams for supporting communication, the orientation byPeircès Prag-matism, and the aim of representing and processing knowledge. These concerns open rich possibilities of interplay and integration. We discuss the philosophical foundations of both disciplines, and analyze their speciic qualities. Based on this analysis, we discuss some possible approaches of interplay and integration. Conceptual structures in knowledge representation are models (or artifacts) representing a perceived reality. With computer applications, these models support and delimit the kind of processing that a knowledge system will be able to carry out. Hence, the semantics they carry is of the outmost importance to the sound-ness and completeness of the applications that use them. With Artiicial Intelligence related applications, these models represent human knowledge in a format available to an inference engine. To describe knowledge, a multitude of knowledge representation formalisms were devised. Among them, semantic networks aim at bringing together knowledge representation and graphical formalisms, hoping to alleviate the knowledge modeling and transfer problems pertaining to knowledge acquisition. Through the years, diierent semantic network based formalisms were introduced , each having a diierent scope. Some dealt with linguistic knowledge, using the assumption that knowledge is necessarily expressible using natural language (cf. 33]). One of the underlying reasons for using this assumption was to facilitate the interfacing between the knowledge system and both the knowledge engineer and the end-user, since both could use a natural language based interface to communicate with the system. Furthermore, this would improve the interpretability of the achievements of the system at any given time. Some other formalisms had preoccupations pertaining to complexity, as their applicability was often challenged by complexity results which discarded them