A Three Level Classification of Entities for Knowledge Representation Systems
In this paper, we present a hybrid classification method for entities in knowledge representation systems including lexical systems. The method introduced is used in the system LINAS (A Natural Language Interface to Bibliographic Databases - Literaturrecherche in natürlicher Sprache) for the specification of lexical entries as well as for the definition of semantic relations in the knowledge representation system, for the support of inferences and for answer generation. On the top level, entities are divided into disjunctive sorts which are philosophically motivated. This level is called epistemic-ontological. In the second level, these sorted entities may be further specified by a given set of features. Because of the relevance of these features for the characterization of human social and cultural activities in the broadest sense, this level is called socio-cultural. Even more specific subcategorization information is provided in the third level, the so-called linguistic subtlety level, which is used for linguistic fine differentiation exclusively. This is done by connecting lexical entities to parts of a semantic network. So a fine grained subcategorization is possible and a continuous transition from lexical knowledge to world knowledge can be achieved.
Nutzung und Vervielfältigung:
Alle Rechte vorbehalten