성과공개

4차

구성기술3

경희대학교

학술발표

설계 적법성 평가를 위한 대규모언어모델 활용 공간 객체의 법규 속성 추론

한국CDE학회 2024 동계학술대회 논문집

이세진

김인한

김동영; 김태원

'Since the emergence of ChatGPT, generative AI has been developing rapidly. In the construction industry, various cases of utilizing large language models, which are a field of generative AI, are being attempted. Large language models, including ChatGPT, show excellent effectiveness in understanding and analyzing natural language, and can generate simple program codes or make inferences. In automated rule checking based on BIM, various legal attributes must be entered into each object. In particular, in the case of space, various attributes must be included along with the space classification, but the types of attributes are different and the checking methods are different, resulting in confusion. In this paper, the authors present a method to automatically infer spatial attribute information using ChatGPT, a large language model. When a BIM model is loaded within the automated rule checking system, ChatGPT is mapped and analyzed based on information entered in advance and information extracted from the model, and attribute information is returned. Through this study, designers are expected to be able to prevent duplicate attributes and apply a variety of design expression methods.