First-Order Intuitionistic Linear Logic and Hypergraph Languages

TL;DR

This paper explores the relationship between first-order intuitionistic linear logic and hypergraph languages, introducing a new hypergraph logic grammar framework and analyzing its generative power and semantics.

Contribution

It introduces hypergraph first-order logic categorial grammars, demonstrating their generative capabilities and developing hypergraph language models for MILL1 with completeness results.

Findings

Hypergraph ILL1 grammars generate all recursively enumerable hypergraph languages.

Hypergraph MILL1 grammars are as powerful as linear-time hypergraph transformation systems.

String MILL1 grammars are closed under intersection and include complex languages.

Abstract

The Lambek calculus is a substructural logic known to be closely related to the formal language theory: on the one hand, it is used for generating formal languages by means of categorial grammars and, on the other hand, it has formal language semantics, with respect to which it is sound and complete. This paper studies a similar relation between first-order intuitionistic linear logic ILL1 along with its multiplicative fragment MILL1 on the one hand and the hypergraph grammar theory on the other. In the first part, we introduce a novel concept of hypergraph first-order logic categorial grammar, which is a generalisation of string MILL1 grammars studied e.g. in Richard Moot's 2014 works. We prove that hypergraph ILL1 grammars generate all recursively enumerable hypergraph languages and that hypergraph MILL1 grammars are as powerful as linear-time hypergraph transformation systems. In addition, we show that the class of languages generated by string MILL1 grammars is closed under intersection and that it includes a non-semilinear language as well as an NP-complete one. This shows how much more powerful string MILL1 grammars are as compared to Lambek categorial grammars. In the second part, we develop hypergraph language models for MILL1. In such models, formulae of the logic are interpreted as hypergraph languages and multiplicative conjunction is interpreted using parallel composition, which is one of the operations of HR-algebras introduced by Courcelle. We prove completeness of the universal-implicative fragment of MILL1 with respect to these models and thus present a new kind of semantics for a fragment of first-order linear logic.