Replication data for: Digital Newspaper Detection System for Interest Groups Strategies in the European Union

This repository contains all data produced for the Germany case study. Using STS techniques and computational text analysis, it identifies communication strategies based on alignment with legislative and general objectives of German interest groups listed in the Transparency Register. The corpus covers German-language digital news outlets across the European Union and the United Kingdom. Data reflect a snapshot as of 16 August 2025. This are the files structures. Analisis_con_progreso_de_20250816.parquet Grain: one row = one German-language article with one of the Groups of interest, combinating all of them. Fields: tipo: record type (e.g., news). texto_comparado: processed article text/fragment used for the similarity matching. similitud_objetivos: similarity to the group’s general goals (0–1). similitud_legislacion: similarity to the group’s legislative targets (0–1). similitud_combinada: composite similarity of the two above (0–1). idioma: language code (de). metadata: JSON with article metadata; url title date Newspaper Search_Category: Transparency Register typology of the group. agenda_category: alignment label under the study’s rules (Lobby & Policy aligned | Agenda Goals | Policy driven | No evidence). outlet: normalized outlet identifier. ab_articles_de_2025-08-16.csv Grain: one row = one German-language article with the best-aligned German interest group for that article. Fields: tipo: record type (e.g., news). grupo_id: EU Transparency Register ID (format ##########-##). grupo_nombre: official name of the interest group. texto_comparado: processed text/fragment used for the similarity computation. similitud_objetivos: article ↔ general goals similarity (0–1). similitud_legislacion: article ↔ legislative targets similarity (0–1). similitud_combinada: combined similarity (0–1). idioma: language code (de). metadata: JSON with article metadata (url, titulo, fecha dd/mm/yyyy HH:MM:SS, periodico). agenda_category: final category assigned to the article (see Glossary). outlet: normalized outlet identifier. goals_ge_thr: TRUE if similitud_objetivos ≥ threshold. legis_ge_thr: TRUE if similitud_legislacion ≥ threshold. joint_min: min(similitud_objetivos, similitud_legislacion) (0–1). cat_lobby_policy_aligned, cat_agenda_goals, cat_policy_driven, cat_no_evidence: mutually exclusive boolean flags of the final category.Notes: decision threshold used in the study = 0.65. If neither similarity reaches the threshold, the article is “No evidence”. ab_alignment_by_outlet_de_2025-08-16.csv Grain: one row = one outlet. Fields: outlet: outlet name/ID. n_obs: number of analyzed articles for the outlet. pct_aligned: percent of “Lobby & Policy aligned”. pct_non_aligned: percent not aligned (pct_goals + pct_policy + pct_none). pct_goals: percent “Agenda Goals”. pct_policy: percent “Policy driven”. pct_none: percent “No evidence”.Notes: percentages are in 0–100. ab_top10_outlets_de_2025-08-16.csv Grain: top-10 outlets by highest pct_aligned. Schema: same columns and definitions as ab_alignment_by_outlet_de_2025-08-16.csv. ab_bottom10_outlets_de_2025-08-16.csv Grain: bottom-10 outlets by lowest pct_aligned. Schema: same columns and definitions as ab_alignment_by_outlet_de_2025-08-16.csv. ab_categories_de_2025-08-16.csv Grain: one row = one global alignment category. Fields: category: one of {Lobby & Policy aligned, Agenda Goals, Policy driven, No evidence}. count: number of articles in the category. perc: percentage over the total (0–100). ab_by_typology_counts_de_2025-08-16.csv Grain: one row = one Transparency Register typology. Fields: Search_Category: typology (e.g., Companies & groups, NGOs, Law firms, Academic institutions, etc.). Lobby & Policy aligned: absolute count. Agenda Goals: absolute count. Policy driven: absolute count. No evidence: absolute count.Notes: all values are counts. ab_by_typology_perc_de_2025-08-16.csv Grain: one row = one Transparency Register typology. Fields: same column names as the counts file, but each cell is the percentage (0–100) of articles in that typology falling into each category. ab_by_typology_combined_de_2025-08-16.csv Grain: one row = one Transparency Register typology. Fields: Search_Category: typology. count_Lobby & Policy aligned, count_Agenda Goals, count_Policy driven, count_No evidence: absolute counts by category. perc_Lobby & Policy aligned, perc_Agenda Goals, perc_Policy driven, perc_No evidence: percentages by category (0–100). n_total_typology: total articles in that typology. ab_by_typology_stats_de_2025-08-16.csv Grain: global association statistics between typology and alignment categories. Fields: chi2: chi-square statistic. p_value: p-value. dof: degrees of freedom. cramers_v: effect size (0–1). n_total: total observations. r_levels: number of typology levels. k_levels: number of categories (4). ab_joint_stats_de_2025-08-16.csv Grain: global summary of joint similarity across all article–group comparisons. Fields: n_total: total number of article–group comparisons used. n_outlets: number of distinct outlets. joint_mean: mean of joint_min (0–1). q10, q25, q50, q75, q90: quantiles of joint_min. share_goals_ge_thr: share (0–1) with similitud_objetivos ≥ threshold. share_legis_ge_thr: share (0–1) with similitud_legislacion ≥ threshold. share_both_ge_thr: share (0–1) where both similarities ≥ threshold.Notes: shares are proportions (0–1), not percentages. Glossary of final categories Lobby & Policy aligned: the article reaches the threshold on both general goals and legislative targets (strong alignment). Agenda Goals: only the general goals similarity reaches the threshold. Policy driven: only the legislative targets similarity reaches the threshold. No evidence: neither similarity reaches the threshold.