Towards A First Order Logic Representation of Complex Situations in Natural Language Understanding

Abstract

B. O. Akinkunmi
ope34648@yahoo.com
Department of Computer Science,
University of Ibadan, Ibadan, Nigeria

Abstract
In natural language understanding (NLU), situations are identified as well as the causal and temporal relations between them. In
Episodic Logic (EL), a knowledge representation scheme for NLU, both atomic situations (characterized by simple assertions
like “tom greets mary”) and complex situations (characterized by complex assertions such as “everyone greets mary) are allowed
as they naturally arise in narrative texts. The problem of allowing both kinds of situation in EL, is that operators extraneous to
first-order logic are used which interpretation yield statements that treat complex propositional statements as terms the such as:
PC(?x.Greet(x, mary), s) where the predicate PC represents the relation partially-characterizes. This paper develops a formalism
based on the situation calculus that expresses simple and complex situations of the same nature as those in EL without leaving
the confines of first-order logic as needed when using EL. The approach taken here is to use a reified form of McCarthy and
Hayes? situation calculus(SC) as the basis for expressing complex situations. The ability of the reified SC-like formalism to
correctly interpret situations involving complex as well as simple propositions is demonstrated. The achievement of this paper is
a major step towards making FOL an adequate semantic representation language for narrative texts.