Topic modeling of budget laws as policy roadmaps: analyzing economic priorities in Italy

This study investigates the use of topic modeling techniques to analyze the industrial priorities embedded within Italian budget laws from 2020 to 2023. By employing four methods -Latent Semantic Analysis (LSA), Fuzzy Latent Semantic Analysis (fLSA), Latent Dirichlet Allocation (LDA), and Correlated Topic Model (CTM)- the analysis identifies coherent expenditure topics and assesses their relevance to fiscal and industrial strategies. Results indicate that, within the corpus examined here, fLSA and CTM provide more informative representations than LSA and LDA in terms of stability and complexity preservation, with fLSA offering more parsimonious solutions and CTM capturing more nuanced topic correlations. The findings illustrate how a comparative use of topic-modeling approaches can support the interpretation of policy priorities in legal and budgetary texts.