Geometry of language

Abstract

Girard (1987) introduced proof nets as a syntax of linear proofs which eliminates inessential rule ordering manifested by sequent calculus. Proof nets adapted to the Lambek calculus (Roorda 1991) fulfill a role in categorial grammar analogous to that of phrase structure trees in CFG so that categorial proof nets have a central part to play in computational syntax and semantics; in particular they allow a reinterpretation of the "problem" of spurious ambiguity as an opportunity for parallelism. This article aims to make three contributions: i) provide a tutorial overview of categorial proof nets, ii) apply and provide motivation for proof nets by showing how a partial execution eschews the need for semantic evaluation in language processing, and iii) analyse the intrinsic geometry of partially commutative proof nets for the kinds of discontinuity attested in language, offering proof nets for the in situ binder type-constructor Q(., ., .) of Moortgat (1991/6).