Metro construction is a typical large-scale complex high-risk engineering project, and construction safety risk management is knowledge-intensive. For the safety risk management of metro construction, the value of knowledge lies in the ability to support people to complete specific risk management tasks. This knowledge is called the domain knowledge of metro safety risk management. However, the current knowledge of metro construction safety risk management is not clear, and people do not know which domain knowledge to master in order to complete safety risk management. Therefore, this paper aims to build the domain knowledge of metro construction safety risk management based on natural text processing and data mining techniques from a large number of metro construction safety texts.Four analysis processes of domain knowledge construction are proposed: domain knowledge extraction, domain knowledge extension, domain knowledge hierarchy extraction and domain knowledge testing. The domain knowledge extraction aims to extract the knowledge gap in the process of metro construction safety risk management. The extraction rules consist of specific part-of-speech combinations, specific dependencies, and specific word lengths. Through the text mining of 158 accident investigation reports, a total of 1739 domain knowledge was extracted, which was reduced to 188 after synonym determination and standardization. The exacted domain knowledge is usually a noun (n) or a noun phrase between 2-6 words.The domain knowledge extension aims to expand the domain knowledge of the former step. Using the Co-occurrence, Support, and Lift indicators in the association rules, 68,816 related documents were analyzed. Finally, the number of domain knowledge was expanded from 188 to 1583.Domain knowledge hierarchy extraction is designed to achieve the extraction of hierarchical relationships between domain knowledge. Using natural language processing techniques, according to the specific syntactic structure and relationship indicators existing among the domain knowledge, the homotopic relationship and the upper and lower relationship between the domain knowledge are determined. The preliminary structure of domain knowledge of the metro construction safety risk management is constructed through the domain knowledge hierarchy extraction process.Domain knowledge testing aims to test the domain knowledge and domain knowledge structure that has been proposed. A total of 119 pre-survey and 425 formal survey questionnaires were issued. Drawing on statistically relevant indicators, including reliability, validity, skewness, peak value, test statistic, and rejection domain, the domain knowledge structure was finally determined. The domain knowledge consists of 4 major categories, 11 subcategories.This paper applies knowledge theory to the field of metro construction safety risk management. By studying the domain knowledge and structure of metro construction safety risk management, it lays a foundation for the continuous accumulation, storage and reuse of knowledge. The domain knowledge analysis method proposed in this paper is not only suitable for analyzing the knowledge of metro construction safety risk management, but also for other types of engineering projects and other areas of knowledge analysis
