A Framework to Analyze the Coordination of Policy Objectives Using Topic Modeling

Abstract

The effective implementation of policies lies in coordinating objectives across different governments. Latent Dirichlet Allocation (LDA) models are crucial for analyzing the thematic content of policy texts. However, a single LDA model struggles to capture the richness of policy topics and content from different policy actors, and its output consists of probabilistic word clusters, which significantly reduces interpretability. These issues result in ineffective policy coordination objective measurement models. We first constructed multi-LDA model to retain the unique characteristics of different policy actors, and developed a policy objective framework to facilitate an effective association between LDA topics and policy objectives. Finally, semantic information was captured using the Word2Vec model, measuring policy objective coordination through a “ word - topic - document” hierarchy. Analyzing 112 Chinese New Energy Vehicles (NEVs) policy documents from 2009 to 2022, this approach assessed the central-local coordination and provided both theoretical and practical insights for policy implementation.