Construction of a Multi-Dimensional Policy Retrieval Model for Personalized Needs

This paper addresses the critical limitations of existing policy information systems that fail to provide personalized and context-aware retrieval for diverse user communities. We propose a novel multi-dimensional policy retrieval framework that enhances traditional BM25 algorithms with context-aware weighting mechanisms, incorporating geographic, sectoral, behavioral, and temporal dimensions to capture the multifaceted nature of policy relevance. The framework employs sophisticated user modeling techniques and advanced Chinese text processing methods to deliver contextualized policy recommendations. Experimental validation on 150,000 Chinese government policy documents demonstrates substantial improvements across all evaluation metrics, with Precision@10 reaching $\mathbf{8 7. 3 \%}$, MAP achieving $\mathbf{7 2. 1 \%}$, and user satisfaction scores increasing by 42% compared to baseline methods. The proposed approach significantly advances personalized information retrieval theory while providing practical solutions for improved policy accessibility across enterprises and organizations, establishing foundations for digital governance transformation through intelligent policy information systems that support more transparent and effective governance paradigms.