A framework for analyzing water policy evolution was constructed based on machine learning, text measurement, and knowledge graph. With 252 water policy documents issued in Wuhan City from 2001 to 2021 taken as research samples, a keyword dictionary construction technique based on the Word2Vec model and a keyword hotness calculation method were proposed to explore water policy evolution laws of Wuhan City from two aspects of policy Keywords and policy instrument. The results show that water policy evolution in Wuhan City experienced three stages during the study period, including initial exploration, comprehensive management, and ecological priority, and a relatively complete water policy system was initially formed. The focus of water policy gradually shifted from water supply and demand to ecological civilization construction. Policy instruments were dominated by the type of command and control, and the economic incentive policy instruments demonstrated a decreasing trend, while the proportions of the technology-supporting policy instruments and public participation policy instruments increased. Water policy development is currently facing problems such as a relatively weak concept of systemic governance, insufficient attention to flood prevention and control, and an imbalance in types and structures of policy instruments. In the future, the water policy system can be improved from aspects of adhering to systemic water management, enhancing flood control and drainage capacity, and improving the structure of policy instruments.