Towards a Computational Formalization of Natural Language Semantics

\ I~I~RODUCTION The problem of explicating the meaning of natural language discourse and how this meaning is obtained has been of central importance to the study of philosophy since the time of Aristotle. This problem has continued to perplex philosophers, linguists, psychologists, and other academicians right up to this very day, such perplexity manifesting itself in disputes such as nominalism vs. realism, behaviorism vs. mentalism, logical positivism vs. the "ordinary-language" approach, and structuralism vs. trans-formationalism. Basic to this perplexity has been the lack of an adequate set of tools with which to formalize meaning in natural language, resulting in a basic split between those who use existing tools to produce formalizations that are oversimplified and inadequate and those who attempt to account for the full complexity of natural language but, in doing so, abandon any attempt at formalization. Clearly, any move toward providing a more powerful set of tools with which to formalize the semantics of natural languages will have implications that will reverberate through a variety of academic disciplines. A formalization of natural language semantics that is keyed to the computer will have a variety of very practical applications as well. Two decades of research in machine translation have failed to produce effective systems, chiefly because machines have not yet been able to duplicate the translator's function of understanding the material in the source language and then restating it in the target language. Once computers are capable of analyzing natural language to a depth where it becomes possible to mechanically determine equivalence of meaning in context of statements in two different languages, and capable of generating well-formed natural-language discourse from such an analysis, mechanical translation of scientific and other expository