基于多维度和动态图神经网络的单据智能审核方法及系统
By constructing an intelligent document review method based on multi-dimensional and dynamic graph neural networks, the accuracy and efficiency problems of document review in existing technologies are solved. It realizes credit-aware review, dynamic rule collaboration and spatiotemporal risk fusion, thereby improving the accuracy and efficiency of document review.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INSPUR GENERSOFT CO LTD
- Filing Date
- 2026-04-15
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies cannot effectively capture the hierarchical relationship and cross-field dependencies between the main table and sub-tables in business document review. They are difficult to adapt to dynamic changes in business strategies, lack automatic optimization capabilities based on data feedback, have low risk assessment coverage in complex scenarios, and are not sufficiently coupled with credit status and review strategies, resulting in low review efficiency.
A document intelligent review method based on multi-dimensional and dynamic graph neural networks is adopted, and a three-level collaborative architecture of credit assessment, graph neural network parsing and spatiotemporal risk prediction is constructed. Through graph structure parser and multi-dimensional spatiotemporal risk prediction model, credit rating is calculated in real time and user credit, graph features and spatiotemporal risk factors are integrated, and feature weights are dynamically adjusted for review.
It improves the accuracy and efficiency of document review, reduces the review time for high-credit users, enhances the accuracy of complex structure parsing and risk identification coverage, and achieves self-optimizing response of rules and neural networks to adapt to dynamic environments and user behavior.
Smart Images
Figure CN122415012A_ABST