FunctionalTypeTheory.docVIP

CHAPTER TWO: FUNCTIONAL TYPE THEORY 2.1. Compositionality We have given a compositional semantics for predicate logic, a semantics in which the meaning of any complex expression is a function of the meanings of its parts and the way these meanings are put together. In predicate logic, complex expressions are formulas, and their parts are formulas, terms (constants or variables), n-place predicates, and the logical constants. As is well known to anybody who has gone through lists of exercises of the form: translate into predicate logic, predicate logic does a much better job on representing the meanings of various types of natural language sentences, than it does on representing the meanings of their parts. Take for example the following sentence: (1) Some old man and every boy kissed and hugged Mary. With the one-place predicate constants MAN, OLD, BOY and the two-place predicate constants KISSED and HUGGED, we can quite adequately represent (1) in predicate logic as (1): (1) (x[MAN(x) ( OLD(x) ( KISSED(x,m) ( HUGGED(x,m)] ( (y[BOY(y) ( KISSED(y,m) ( HUGGED(y,m)] The problem is that without an explicit syntax and a compositional semantics, representing (1) as (1) is a completely creative exercise, and it shouldnt be. What we really need is a semantics that gives us the right semantic representation for sentence (1) (and hence the right meaning), and that tells us at the same time how this semantic representation is related to the meanings of the parts of (1). Our representation (1) has various parts (like () that dont correspond to any part of the sentence; some parts of the representation occur twice in different forms, while in the sentence there is only one part (kissed and hugged Mary); the representation contains sentence conjunction (, whereas the sentence contains NP conjunction and verb conjunction; OLD is separated from MAN in the representation, which is motivated by what the meaning should be, but nothing explains how it got separated, etc. What w