With the brisk advancement of information technology, the significance of digital-driven learning within the educational domain has been growing more and more prominent. Knowledge graphs, functioning as an efficacious instrument for depicting and arranging knowledge, assume a crucial and indispensable role in digital-driven learning. Nevertheless, traditional approaches to knowledge graph construction encounter issues like inefficiency and incomplete knowledge coverage when confronted with large-scale and intricate learning scenarios, rendering it arduous to satisfy the practical requirements. In order to address this challenge, this paper initially presents the data processing and knowledge extraction methodologies grounded in deep learning, along with the deep learning-driven knowledge graph construction and fusion strategies. Subsequently, the deep learning algorithm is employed to extract knowledge and construct graphs from this information, thereby forming a structured representation of knowledge. Ultimately, through the fusion strategy, the knowledge graphs sourced from diverse origins are effectively integrated, enhancing the comprehensiveness and precision of the knowledge. The outcomes of the experimental investigation reveal that the utilization of deep learning technology is capable of augmenting the knowledge coverage in the construction of knowledge graphs. The knowledge coverage of the optimized knowledge graph is substantial and consistently approaches 1, offering robust technical support for digital-driven learning. Deep learning technology commences by extracting pivotal information from the text via efficient data processing techniques.