Classical linear logic, cobordisms and categorical semantics of categorial grammars

Sergey Slavnov
DOI: https://doi.org/10.48550/arXiv.1810.02047
2019-02-10
Abstract:We propose a categorial grammar based on classical multiplicative linear logic. This can be seen as an extension of abstract categorial grammars (ACG) and is at least as expressive. However, constituents of {\it linear logic grammars (LLG)} are not abstract ${\lambda}$-terms, but simply tuples of words with labeled endpoints, we call them {\it multiwords}. At least, this gives a concrete and intuitive representation of ACG. A key observation is that the class of multiwords has a fundamental algebraic structure. Namely, multiwords can be organized in a category, very similar to the category of topological cobordisms. This category is symmetric monoidal closed and compact closed and thus is a model of linear $\lambda$-calculus and classical linear logic. We think that this category is interesting on its own right. In particular, it might provide categorical representation for other formalisms. On the other hand, many models of language semantics are based on commutative logic or, more generally, on symmetric monoidal closed categories. But the category of {\it word cobordisms} is a category of language elements, which is itself symmetric monoidal closed and independent of any grammar. Thus, it might prove useful in understanding language semantics as well.
Logic in Computer Science,Computation and Language,Logic
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is to propose a categorial grammar based on classical multiplicative linear logic. This new grammar model aims to extend Abstract Categorical Grammars (ACG) and be at least as expressive as it. Specifically, the paper addresses the following aspects: 1. **Providing an Intuitive Representation**: Unlike ACG, the newly proposed Linear Logic Grammars (LLG) use multiwords as basic language elements instead of abstract λ - terms. A multiword is a tuple of words with marked endpoints, which makes them more easily recognizable as basic components of a language. 2. **Identification of Algebraic Structures**: The paper points out that the multiword class has a fundamental algebraic structure and can be organized into a category, similar to the category of topological cobordisms. This category is symmetric monoidal closed and compact - closed, and thus is a model of linear λ - calculus and classical linear logic. This finding not only provides a theoretical basis for LLG but may also provide a categorical representation for other formal systems. 3. **Understanding Language Semantics**: Many language semantics models are based on commutative logic or symmetric monoidal closed categories. And the category of word cobordisms itself is a symmetric monoidal closed category of language elements independent of any grammar. Therefore, it helps to understand language semantics and may establish connections between different grammar systems. ### Formula Representation To ensure the correctness and readability of the formulas, the following are some key formulas involved in the paper and their Markdown - format representations: - **Tensor Product**: \[ X \otimes Y \] - **Monoidal Unit**: \[ 1 = 1_r = 1_l = \emptyset \] - **Internal Hom**: \[ X \multimap Y = X^\perp \otimes Y \] - **Duality**: \[ X^\perp = X, \quad (X^\perp)_r = X_l, \quad (X^\perp)_l = X_r \] - **Natural Pairing Map**: \[ \epsilon_A : A \otimes A^\perp \to 1 \] Through these improvements, the paper not only extends the capabilities of existing grammar models but also provides a more intuitive and more algebraically - structured way to understand and handle complex relationships in language.