Toward Domain-Specific Large Language Models in Construction Engineering: A Review of Development Tools

Abstract The architecture, engineering, and construction industry is facing the challenge of massive heterogeneous data, with an urgent demand for domain-specific large language models (LLM) capable of understanding complex engineering semantics and meeting high reliability requirements. Focusing on the four existing approaches for the specialization of general purpose LLM specialization, this paper aims to systematically review the development tools for domain-specific LLM specialization in construction engineering based on general-purpose LLM specialization. The development tools are classified into three categories: data preparation; basic LLM specialization; and advanced LLM specialization. Subsequently, their features are summarized with typical examples. Finally, the selection strategies are recommended, and the current challenges are summarized. This paper enhances the body of knowledge on artificial intelligence applications by offering systematic and practical insights into the development tool ecosystem for domain-specific LLM specialization in construction engineering.