Testing ChatGPT-aided SparQL generation for semantic construction information retrieval

Recently there has been a strong interest in using semantic technologies to improve information management in the construction domain. Ontologies provide a formalized domain knowledge representation that provides a structured information model to facilitate information management issues such as formalization and integration of construction workflow information and data and enables further applications such as information retrieval and reasoning. SPARQL Protocol And RDF Query Language (SPARQL) queries are the main approaches to conduct the information retrieval from the Resource Description Framework (RDF) format data. However, there is a barrier for end users to develop the SPARQL queries, as it requires proficient skills to code them. This challenge hinders the practical application of ontology-based approaches on construction sites. As a generative language model, ChatGPT has already illustrated its capability to process and generate human-like text, including the capability to generate the SPARQL for domain-specific tasks. However, there are no specific tests evaluating and assessing the SPARQL-generating capability of ChatGPT within the construction domain. Therefore, this paper focuses on exploring the usage of ChatGPT with a case of importing the Digital Construction Ontologies (DiCon) and generating SPARQL queries for specific construction workflow information retrieval. We evaluate the generated queries with metrics including syntactical correctness, plausible query structure, and coverage of correct answers