Research on Knowledge Graph Technology with Unbalanced Samples

Abstract The application of the existing knowledge graph is widely concentrated in civil fields with sufficient sample size and uniform distribution, especially Internet, finance other fields. Due to particularity its background military field, there a general lack size, which greatly limits support ability atlas technology for intelligent applications. Therefore, how solve problem automatic extraction multimodal data under condition unbalanced samples key core that affects effect perception field. This topic focuses on characteristics small batch, multi imbalance diversity field operations, introduces self-monitoring method based deep learning, solves low accuracy unstructured data, dependence traditional algorithm carries out research construction atlas, constructs equipment type spectrum library.