Derived Data and Reproduction Code for the Longitudinal Analysis of Policy Attention in South Korea's Smart Farm Innovation Valleys, 2019–June 2026

This repository provides derived data and independently reconstructed reproduction code associated with the study "Longitudinal Analysis of Policy Attention in South Korea's Smart Farm Innovation Valleys," covering the period from 2019 to June 2026. The analytical dataset consists of 15,455 documents retained from an initial TEXTOM export of 24,036 records through documented temporal screening, relevance filtering, and deduplication. Documents were assigned to three analytical phases: 2019–2020 (Phase 1), 2021–2022 (Phase 2), and 2023–June 2026 (Phase 3), using the year-level retrieval-window information recorded in the TEXTOM export rather than independently verified publication dates. The archive includes: A fixed bilingual dictionary of 74 analytical concepts. Documented screening decisions for 141 candidate concepts, including 74 retained and 67 excluded concepts. Phase-specific concept frequencies and 74 × 74 co-occurrence and cosine-similarity matrices. Text-free binary document-by-concept detection matrices covering the 15,455 analytical records. Source-year aggregation matrices and summary results from manual semantic validation. Independently reconstructed verification scripts for quadratic assignment procedures, block bootstrap analyses, fixed-margin randomization, and weighted Louvain community detection. Documented random seeds, reproduction instructions, and SHA-256 file-integrity checks. The materials enable independent verification of the principal descriptive network matrices and selected statistical and relational findings. The supplied verification scripts use explicitly documented random seeds and are distinguished from the original analytical code where the original implementations or random-number streams could not be recovered. Original third-party document texts, titles, URLs, and individual manual-annotation records are not redistributed. Certain components of the original analytical workflow, including parts of the dependence-adjusted modeling and vocabulary-sensitivity procedures, cannot be fully reconstructed from the released materials alone. These limitations are documented in the accompanying README and reproduction-status files. This archive is provided to support research transparency, methodological evaluation, and reproducibility of the specified analyses of documented policy attention and lexical network relationships.