Computational Semantics:A Study of a Class of Verbs

Abstract

This thesis starts by showing that the dominant linguistic theory of the last fifteen years, Transformational Generative Theory, has now been so criticized that it is no longer viable as a Theory of Language. An alternative theory is proposed, called the Functional Theory of Language (FTL). This theory should be extremely useful for the field of Artificial Intelligence, since it would allow computer to understand Natural Language in any context. FTL proceeds from the realization that Language is used to convey information from one person to another. A class of verbs, explicitly performative verbs, is distinguished. Use of such verbs in sentences displays unambiguously the intentions of the speaker (the information he wishes to convey). This information is carried by the presuppositions inherent in the verb, which limit the choice of verb to that which accurately reveals the attitudes of the speaker (his cognitive structure). Two computer programs have been written to test FTL, both of which only accept sentences containing explicitly performative verbs. (It is theoretically possible for any English sentence to be "reduced" to sentences containing explicitly performative verbs.) The first program detects inconsistencies in the speech of various 'people'; that is to say, it tests how well understanding of English can be achieved according to FTL. The second program ‘converses' with the user; that is to say, it tests how well English can be 'generated' according to FTL. The conclusions are drawn that the programs mark a significant advance in the field of Artificial Intelligence and that, on the basis of those programs, the Functional Theory of Language should be considered further by philosophers of Language.