Combinatory Categorial Grammar (CCG) is a theory in which lexical words and derived constituents of natural language have directional functional types---hence the term "Categorial"---and in which the grammatical operations which build derivations are strictly type-driven, rather than structure-driven---hence "Combinatory". Under such a theory, the linguistic notion of "constituent" is considerably generalized: most (but not all) contiguous substrings of a sentence are typable constituents, and all typable constituents have a compositionally assembled semantic interpretation. Important advantages ensue for analyzing various problematic phenomena of coordination, intonation structure, and incremental processing.

This talk concentrates on its application to intonation structure in spoken language processing. Intonation in English is used to convey the distinction within an utterance between the part that corresponds to question, topic or theme that can be assumed to be mutually agreed by the participants, and another part that is the speaker's novel contribution on that that topic. Controlling intonation correctly is crucial to any kind of interactive spoken dialog, and has hitherto proved problematic. Several examples drawn from real human computer interactive applications will be given to show the practical importance of this problem. An analysis will be offered in terms of CCG and there'll be some discussion of the problem of parsing and generation with CCG.