A data grading method and system based on a three-level knowledge base

Abstract

The application discloses a data grading method and system based on a three-level knowledge base, relates to the technical field of data governance, and comprises the following steps: preconstructing a three-level knowledge base, and completing vectorization and weight initialization on the content of each level; when a grading task is triggered, performing progressive weighted retrieval on a field to be graded based on the preconstructed three-level knowledge base; assembling an inference input package based on the weighted retrieval result and inputting the inference input package into a large language model to perform grading inference, and outputting a grading result; sending the grading result to an artificial auditing process, and updating the weight parameters and instance entries of the corresponding level of the three-level knowledge base based on the artificial auditing result, so that closed-loop iterative optimization of the knowledge base is realized. The application adopts three-level knowledge base progressive weighted retrieval and layered constraint inference, strengthens the anchoring effect of compliance basis, guarantees the quantifiable traceability of the grading result, and takes into account the grading accuracy and execution efficiency.