Knowledge Graphs And Logic: One Of Two Kinds
Harmen Van Den Berg
Institution: Universiteit Twente (THE
); 0237 NETHERLANDS
Source: DAI, 55, no. 03C, (1993): 0940
Standard No: ISBN: 90-9006360-9
Many current approaches of natural language understanding are objectivistic, which means that the semantics of natural language is described independently of the human being engaged in the process of understanding. As a consequence, objectivist theories reduce the semantics of natural language to truth conditions. These approaches often use logic-like formal languages to describe the semantics of language.
In the first part of this thesis a model for natural language understanding is developed that, unlike most current approaches, reflects the important role humans play in determining the meaning of language. The formal language used is Knowledge Graphs. Knowledge graphs are networks of concepts and relations, where the relations are chosen from a limited set of relation types. The meaning of a natural language sentence is defined as the knowledge graph obtained by combining the graphs that are associated with the words that sentence consists of.
In the second part of this thesis three linguistic phenomena closely connected to logic are discussed, namely negation, quantification, and modality. It is shown that, by introducing negative and modal contexts, these phenomena can be adequately treated. In this way also donkey anaphora can be satisfactory dealt with. The introduction of these contexts amounts to extending the system of existential graphs of C. S. Peirce, yielding logical systems in terms of graphs which are equivalent to propositional logic, predicate logic, and the modal propositional logics K, T, S4, S5, K45, and KD45, respectively. In this way, it is possible to perform logical inferencing without being forced into an objectivistic framework.
Descriptor: COMPUTER SCIENCE
Accession No: AAGC358868