联系我们

论文查询结果	返回搜索

论文编号：	13188
作者编号：	1120180970
上传时间：	2022/6/7 11:17:43
中文题目：	基于语义建模的我国大数据政策扩散特征及其驱动机制研究
英文题目：	Research on the characteristics and driving mechanism of big data policy diffusion in China based on semantic modeling
指导老师：	王芳
中文关键字：	政策扩散；特征分析；语义建模；驱动机制；因果分析
英文关键字：	Policy diffusion; Characteristic analysis; Semantic modeling; Driving mechanism; Causal analysis
中文摘要：	近年来，随着全球化和信息化进程的加快，不同国家和地区间的公共政策扩散现象愈加频繁。高效、健康的政策扩散对于降低政策风险、贯彻统一部署以及提升整体性政府治理水平发挥重要作用。但目前，政策扩散中存在着跨层级政府交流欠缺、扩散水平区域差异大以及政策扩散主体协同性不足等问题，一定程度上制约了政策的实施效果。同时，在政策扩散研究中，针对跨地域、跨层级、跨机构的政策扩散特征及其驱动机制缺乏有效分析，对政策扩散的量化表示与数字化建模存在一定不足。 为探索公共政策扩散分析中的新方式、新模式，本研究以大数据领域的公共政策研究为问题导向，综合运用情报学、公共管理、计算机技术等交叉学科的理论与方法，从语义建模视角下对政策文本进行语义自动化建模分析，对全生命周期内的政策扩散特征（包括时序特征、空间特征、扩散主体特征）进行揭示，深刻剖析了政策扩散与发展背后的动力机制。本研究工作主要包括： 1、对当前政策扩散中存在问题及相关研究进行分析。 2、利用智能化手段对大规模政策文本进行语义建模。构建政策文本数据的语义建模分析框架，提出基于融合BERT词嵌入与BTM模型的政策文本语义建模方法，实现了对政策文本的语义化表达和高维度主题抽取，为后续政策扩散特征分析提供技术支撑。 3、基于政策过程理论，揭示政策扩散时序特征并构建时序特征量化指标模型，揭示不同类型的政策扩散主体在时序上的变化特征。 4、基于政策网络理论，构建基于多维协同和网络测度的政策扩散主体指标体系。设计公共政策扩散主体分析的群落与山峰图可视化方案，剖析政策扩散过程中部分地区所存在的不足之处，为优化政策部署提供情报支撑。 5、基于政策多源流理论，提出融合扩散特征的政策驱动机制研究方法并基于倾向性得分匹配方法对政策扩散因果效能进行量化。探究政策扩散背后的影响因素，利用反事实框架因果分析模型并结合核函数算法、卡尺匹配算法及KNN最近邻算法等，优化政策扩散分组间的偏差度并对政策扩散因果效应进行量化分析。 本研究的主要创新和贡献体现在两个方面：第一，知其然，语义建模视角下对我国公共政策扩散特征进行揭示。利用深度学习和文本挖掘技术对大规模政策信息有效抽取，创新了传统政策分析的研究范式和研究方法，并从扩散时序特征、扩散空间特征、扩散主体特征三个维度，对全国范围内的大数据政策进行扩散特征分析。第二，知其所以然，本研究提出融合扩散特征与反事实因果框架的政策扩散因果分析模型，对政策扩散背后的动力机制进行有效揭示，并利用核函数算法等三种算法进行对比分析，探索了适用于本研究的因果效能量化模型，为有效破解政策治理难题提供技术路径与方法参考。 本研究综合运用情报学、计算机科学、公共管理等多学科理论与方法，探索了如何根据政策文本内部特征与外部属性，基于语义建模分析政策扩散的多维特征要素。理论上，拓展了政策扩散的研究范式，丰富了政策扩散跨学科理论体系。方法上，深入政策文本语义建模，弥补传统单纯数理统计方法的不足。实践上，为政策决策者和宏观分析人员提供整体视角下全生命周期的政策分析。 图49幅，表37个，参考文献258篇。
英文摘要：	In recent years, with the acceleration of globalization and informatization, the phenomenon of public policy diffusion between different countries and regions has become more and more frequent. Efficient and healthy policy diffusion plays an important role in reducing policy risks, implementing unified deployment and improving the overall level of government governance. However, at present, there are some problems in policy diffusion, such as lack of cross level government communication, large regional differences in diffusion level and insufficient synergy of policy diffusion subjects, which restrict the implementation effect of policy to a certain extent. At the same time, in the research of policy diffusion, there is a lack of effective analysis on the characteristics and driving mechanism of policy diffusion across regions, levels and institutions, and there are some deficiencies in the quantitative representation and digital modeling of policy diffusion. In order to explore new ways and models in public policy diffusion analysis, this study takes the public policy research in the field of big data as the problem orientation, and comprehensively uses the theories and methods of interdisciplinary disciplines such as Informatics, public management and computer technology to conduct Semantic Automatic Modeling and analysis of policy texts from the Perspective of semantic modeling, This paper reveals the characteristics of policy diffusion (including temporal characteristics, spatial characteristics and diffusion subject characteristics) in the whole life cycle, and deeply analyzes the dynamic mechanism behind policy diffusion and development. This research work mainly includes: 1.Analyze the existing problems and related research in the current policy diffusion. 2.Semantic modeling of large-scale policy texts by intelligent means. This paper constructs the semantic modeling and analysis framework of policy text data, and puts forward the semantic modeling method of policy text based on the integration of Bert word embedding and BTM model, which realizes the semantic expression of policy text and high-dimensional topic extraction, and provides technical support for the subsequent analysis of policy diffusion characteristics. 3.Based on the policy process theory, this paper reveals the temporal characteristics of policy diffusion, constructs the quantitative index model of temporal characteristics, and reveals the temporal change characteristics of different types of policy diffusion topics. 4.Based on the policy network theory, this paper constructs the policy diffusion subject index system based on multi-dimensional coordination and network measurement. Design the visualization scheme of community and peak map for the analysis of public policy diffusion subject, analyze the shortcomings of some areas in the process of policy diffusion, and provide information support for optimizing policy deployment. 5.Based on the policy multi-source flow theory, a research method of policy driving mechanism integrating diffusion characteristics is proposed, and the causal effectiveness of policy diffusion is quantified based on propensity score matching method. This paper explores the influencing factors behind policy diffusion, uses the counterfactual framework causal analysis model, combined with kernel function algorithm, caliper matching algorithm and KNN nearest neighbor algorithm, optimizes the deviation degree between policy diffusion groups, and makes a quantitative analysis of the causal effect of policy diffusion. The main innovation and contribution of this study are reflected in two aspects: first, know what, to reveal the characteristics of China's public policy diffusion from the perspective of semantic modeling. Using deep learning and text mining technology to effectively extract large-scale policy information, this paper innovates the research paradigm and research methods of traditional policy analysis, and carries out semantic modeling and diffusion feature analysis of big data policies nationwide from three dimensions: diffusion temporal characteristics, diffusion spatial characteristics and diffusion subject characteristics. Second, knowing why, this study proposes a causal analysis model of policy diffusion that integrates diffusion characteristics and counterfactual causal framework, effectively reveals the dynamic mechanism behind policy diffusion, compares and analyzes three algorithms such as kernel function algorithm, and explores the quantitative model of causal effectiveness suitable for this study, Effectively solve policy governance problems and provide technical paths and method references. This study comprehensively uses the multi-disciplinary theories and methods of information science, computer science and public management to explore how to analyze the multi-dimensional feature elements of policy diffusion based on semantic modeling according to the internal and external attributes of policy text. Theoretically, it expands the research paradigm of policy diffusion and enriches the interdisciplinary theoretical system of policy diffusion. In terms of methods, we should go deep into the semantic modeling of policy text to make up for the shortcomings of traditional simple mathematical statistics methods. In practice, it provides policy makers and macro analysts with policy analysis in the whole life cycle from a national perspective. 49 pictures, 37 tables, 258 references.
查看全文：	预览  下载(下载需要进行登录)