AI-BASED DIGITAL METHODS FOR CONSTRUCTION DOCUMENTATION EXAMINATION

The article reviews modern approaches to automating the regulatory examination of architectural and construction documentation using artificial intelligence technologies. The study analyzes the primary tasks involved in the examination of design solutions, specifically the verification of compliance between design parameters and building codes and requirements. Methods for processing textual and graphical information are described, including NLP (Natural Language Processing), computer vision (technical drawing analysis), formal ontologies and knowledge graphs, as well as machine learning and Large Language Models (LLMs). The paper provides an overview of existing solutions and systems that utilize the classification of regulatory documents [1], neural network-based identification of BIM model elements [2], and contextual semantic analysis and rule generation [3]. A comparative analysis of these approaches is performed, identifying their strengths and weaknesses within the context of expert examination. The study discusses current challenges (diversity and complexity of regulations, source data quality, and the need for adaptation to the Russian language) and future prospects (integration of LLMs with BIM, increased digitalization, and data standardization). It is demonstrated that the implementation of AI solutions significantly accelerates project verification and reduces the volume of routine operations for experts [4]; however, it requires further development of methodologies for verification and ensuring the reliability of conclusions. The presented work serves as a reference for researchers and practitioners in the development and implementation of digital systems for design documentation examination.