Abstract: This paper discusses the application of text mining technology in the digital construction of supervision, aiming to solve the problem of information processing in the quality and safety management of housing construction projects. At present, supervisors are faced with a large amount of unstructured text data at the construction site. Although these data contain rich empirical information, they are difficult to be effectively utilized due to their lack of systematic collation. To this end, this study proposes a standardized problem description system based on text mining. Combined with real-time information processing technology, the efficiency and quality of supervision services are improved. The research goal is divided into two parts: to build a quality and safety problem description standardization system. Through text clustering algorithms (such as K-Means), information mining of supervision text data is carried out to form a standard description with clear logic and clear categories. Establish a real-time information processing chain, and use voice input, image recognition and other technologies to instantly obtain on-site problems. The standardized solution is generated by searching the enhanced generation (RAG) model to quickly match the specification provisions. The significance of this study is to transform non-public supervision experience into structured knowledge, and assist supervisors to quickly identify and deal with detailed problems, so as to improve the scientificity and efficiency of project management.