Abstract: The collection and analysis of global strategic mineral resource intelligence is crucial for establishing monitoring and early warning mechanisms. Benefiting from rapid Internet development, governments, mining enterprises, and research institutions have released massive mineral resource information online publicly. However, such information exhibits distinct big data characteristics that pose significant challenges for perception analysis. This paper investigates big data methodologies and information perception models for mineral resource intelligence analysis, constructing an integrated “collection-processing-storage” technical framework for global strategic mineral resource information perception. This framework resolves critical issues in real-time monitoring and data acquisition for global mineral resource analysis. Focusing on key information nodes throughout the lifecycle of strategic mineral resources, it specifically propose four perception tasks: entity recognition, relation extraction, event extraction, and text classification. A 10 000-entry mineral resource perception dataset is constructed, and a BERT-based information perception model is developed to address the structured extraction of semantic content from unstructured mineral resource data. Experimental results demonstrate that the model achieves precision and recall rates exceeding 0.75 across all perception tasks, with an average score across tasks surpassing 0.80. This confirms the model’s strong generalization capability in comprehending domain-specific terminology and accurately capturing elemental correlations, indicating high practical value. This paper presents an automated information perception technical methodology and model for global strategic mineral resources, providing critical information sensing support for early warning systems of strategic mineral resource security.