GDELT global-news theme attention (agent-curated)
Daily enrichment signals scoring which themes dominate each country's news. Each day the 96 keyless GDELT 2.1 GKG fifteen-minute bulk files for the last complete UTC day are downloaded; articles are deduplicated on URL, attributed to one country (first country-level V2ENHANCEDLOCATIONS entry, FIPS 10-4 code resolved to ISO alpha-3), and their V1THEMES topic codes are aggregated per (country, theme) with GDELT's own per-article V1.5TONE scores. Per country-theme cell: n_articles, n_sources, mean_tone, pos_share / neg_share (fraction of articles with tone >= 1.0 / <= -1.0), theme_share_pct (cell articles over the country's themed articles; shares sum above 100 because articles carry several themes), theme_rank (per country, n_articles desc, theme asc), attention_score = 100 * (50% global min-max of n_articles + 50% global min-max of theme_share_pct; 100 = the day's most-attended cell worldwide), and theme_concentration (per-country Herfindahl of theme shares, 0-10000; high = a few themes dominate). Who joins this: a media analyst joins daily country theme attention to the gdelt event/conflict tables on country_code + as_of; a comms team tracks a theme's salience per market. Primary key: (as_of, country_code, theme); join keys: country_code (ISO alpha-3), as_of, theme. Nullability: none — every column is populated for every kept cell.