A comprehensive conceptual analysis using ER and conceptual graphs
Titel:
A comprehensive conceptual analysis using ER and conceptual graphs
Auteur:
Gray, Linda C. Bonnell, Ronald D.
Verschenen in:
Journal of experimental & theoretical artificial intelligence
Paginering:
Jaargang 4 (1992) nr. 2 pagina's 95-106
Jaar:
1992-04-01
Inhoud:
This paper compares two representations used for concept modeling: the entity-relationship (ER) approach, and conceptual graphs (CG), and shows how these techniques can be combined for carrying out the task of conceptual analysis for a domain. The two techniques are contrasted with regard to various semantic aspects: the kind of knowledge being modeled, goals, context, primitives, type of semantics, and type of truth. The conceptual levels of semantic structure and semantic content are defined. Conceptual analysis at the level of semantic structure produces a definition of the scope of domain, and specifies the kinds of associations which can occur in the domain as a whole. At the level of semantic content, conceptual analysis supplies the meanings of terms in more detail, and provides a more fine-grained view of the tokens which can instantiate types. The ER approach is appropriate for modeling concepts at the level of semantic structure, and the CG formalism at the level of semantic content An example of concept modeling and text generation for a simple domain is presented to demonstrate how the ER constructs can be augmented by conceptual graphs. The conceptual analysis for a household survey is developed at both the level of semantic structure and of semantic content. A questionnaire for a household survey is generated from the conceptual analysis, and reflects both the semantic structure of the domain (in the overall outline and layout of the questionnaire), and its semantic content (in the wording of individual questionnaire items). The paper concludes with a discussion of the advantages of simultaneously employing both the ER and CG techniques for modeling the semantics of a domain
Tijdschrift beschrijving