Abstract: In view of problems of lacking intelligent semantic extraction function and multi—attribute interconnection retrieval analysis and decision function，and low intelligence degree in existing intelligent semantic acquisition and decision system of coal mine safety hazards, a kind of intelligent semantic acquisition and smart decision support system of coal mine safety hazards based on improved convolutional neural network (CNN) and ant colony optimization (ACO) was designed. The system adopts CNN—based intelligent semantic acquisition model, and uses CNN algorithm to match the close semantic keywords with the highest similarity, and uses mapping table to concern the standard keywords, so as to solve problem of low matching accuracy of semantic keywords. The system adopts ACO—based intelligent retrieval model, and uses negative feedback and positive reinforcement method of ACO algorithm to mark high—frequency retrieval rules, so as to realize intelligent display of high—frequency retrieval rules. The experiment and application results show that the system can realize functions such as interconnection query of multi—attribute semantic keywords, intelligent display of high—frequency retrieval rules, real—time tracking of data related to hidden danger, and diversified display of data charts, intelligent generation of decision analysis and early warning reports.