Decoding urban policies: NLP-driven concise explanations

Abstract

This study introduces a novel NLP-driven approach for generating accurate explanations of urban policies, addressing the critical need for communication between policymakers and the public. The proposed method integrates policy-specific fine-tuning of large language models, retrieval-augmented generation, and policy-aware prompt engineering. For the policy research, we collect the Zhihu Official Policy Q&A Dataset, a comprehensive collection of 29,151 policy-related questions and answers. Experimental results demonstrate significant improvements in explanation quality, accuracy, and relevance across various policy domains and question types. Human evaluations conducted by urban policy experts and citizens confirm the effectiveness of our method in enhancing the clarity, completeness, and usefulness of policy explanations. The potential implications for urban governance include increased policy transparency, facilitated public participation, and improved policy implementation. While acknowledging limitations such as data bias and model interpretability, this research contributes to the ongoing dialogue on smart city technologies and digital governance, highlighting the potential of NLP-driven approaches to transform urban policy communication.