Review of construction and reasoning methods for knowledge graphs in coal mining domain

Knowledge from diverse data sources in the coal mining domain is extracted to construct a knowledge network. Leveraging reasoning technologies, this network supports equipment fault diagnosis, real-time safety risk warnings and responses, disaster cause analysis, emergency rescue planning, production organization, and operational decision-making, thereby advancing intelligent mining. This paper reviews the research progress on knowledge graphs, with a focus on their applications in coal mining. It discusses the evolution of knowledge-driven artificial intelligence, the architecture of AI systems based on knowledge graphs, primary types of knowledge graphs, and representative studies. The paper examines knowledge modeling, construction, utilization, and maturity of existing knowledge graphs in the coal mining domain. Key challenges in knowledge graph construction and reasoning, spanning entity recognition, relation extraction, graph fusion and error correction, and reasoning, are analyzed. To address these challenges, proposed solutions include span-based entity recognition methods, multi-stack classifier-based relation extraction, entity embedding techniques, and consistency constraint modeling for entity relationships. Research on reasoning techniques should remain application-driven and tightly integrated with business scenarios. Given the abundance of multimodal data such as images and videos in the coal mining field, future efforts could focus on constructing multimodal and temporal knowledge graphs by incorporating time information.