Do Natural Language Processing models understand simulations? Application of GPT-3 to translate simulation source code to English

Abstract

The need for modeling inventory management systems and associated processes is critical. However, the emphasized complexity, along with nonlinear behavior, high dimensionality, and stochasticity, frequently leads to analytic intractability. Simulation modeling does not possess this disadvantage allowing one to describe an inventory management system with all its details and take into account nonlinearity, uncertainty, and complexity. However, simulation modeling has several disadvantages, including technical complexity and the need for the back-and-forth information exchange between the domain experts and simulation engineers, which may significantly increase the project's duration and budget. Our study explores how the latest advancements in Natural Language Processing can be applied to assist in the development of a simulation model of the inventory management system. Our study mainly focuses on the proof of concept that state-of-the-art NLP systems are capable of understanding both the core principles behind the simulations of inventory management systems and the domain-specific context.