Research on the Intent-Driven, Dynamic Pedigree Database Collation Method

Abstract

As the global data volume surges into the Zettabyte era, traditional database organization methods based on predefined schemas encounter significant bottlenecks, such as semantic decoupling, obscure lineage, and prohibitive governance costs when faced with complex business intents and high-frequency data evolution. This paper proposes an innovative paradigm for database organization characterized by “Intent-Driven and Dynamic Pedigree” (IDDP). Firstly, a high-precision intent recognition model is constructed by incorporating an improved Transformer architecture and contrastive learning algorithms, enabling the accurate mapping from natural language intents to logical operational operators. Secondly, a five-tuple dynamic pedigree model comprising Entities, Relations, Time, Versions, and Context is defined to address the challenges of lineage tracking and consistency maintenance during long-cycle data evolution. At the system implementation level, an “Intent-Pedigree-Data” (IPD) three-layer mapping engine and an event-driven incremental update mechanism are designed. Experimental validations conducted on the TPC-DS standard dataset and a real-world e-commerce dataset with tens of millions of records demonstrate that the proposed method achieves an intent parsing accuracy of 89.4% under complex relational queries, representing an improvement of approximately 34% over traditional methods. Furthermore, the dynamic adaptation latency is reduced by 95%, while the pedigree storage overhead is maintained at only 2.4% of the original data volume. This research demonstrates that the IDDP method significantly enhances the autonomous organization capabilities of databases, providing a novel theoretical foundation and engineering pathway for value extraction from large-scale heterogeneous data assets.