Relational automata as a Nexus between artificial intelligence and human linguistic intelligence

This paper introduces a conceptual overview of relational automata that output relations when operating on inputs from the words of a natural language. The central feature of relational automata introduced here is their generation of relations. Crucially, such relational automata build relations in a hierarchical or lattice-like fashion. The specific relations that such automata construct are meaning relations that are constructed on the set of words of a natural language and/or relations defined on words themselves. This article proposes that the construction of meaning relations in more and more complex layers is what relational automata do and this accounts for certain linguistic aspects of meaning making in cognitive systems. The meaning making machinery from a computational perspective can be conceptualized in terms of functioning and outputs of such relational automata, albeit at least in some restricted sense. The novelty of relational automata is that they generate meaning relations, not strings. Also, they have the ability to model intricate natural language phenomena (such as reduplication, discontinuities), with flexibility in handling non-syntactic meaning, which makes it relevant to the AI (Large Language Models) modeling of linguistic meanings. Relational automata formulated in this paper can thus serve to demarcate certain linguistic facets of meaning making in cognitive systems, striking a nexus between AI and human linguistic intelligence. Since the article advances a new model of automata, which link AI systems to competences of human behavior of meaning-making in human language and language comprehension, it turns out that some sort of convergence in semantic representations and computing is possible despite divergence.