GDELT Saudi Arabia Daily News Events

Enriched GDELT 2.0 event records related to Saudi Arabia, updated daily. Each record is deduplicated, geo-enriched with nearest-city matching, and includes scraped article full text when available.

Demo: geonews.tabaqat.dev - Interactive map to explore the data.

Bucket Structure

1s3://us-west-2.opendata.source.coop/tabaqat/gdelt-sa/
2└── country=SA/
3    └── year=YYYY/
4        ├── YYYY_MM_DD.parquet
5        ├── YYYY_MM_DD.parquet
6        └── ...
1s3://us-west-2.opendata.source.coop/tabaqat/gdelt-sa/
2└── country=SA/
3    └── year=YYYY/
4        ├── YYYY_MM_DD.parquet
5        ├── YYYY_MM_DD.parquet
6        └── ...

One Parquet file per day (ZSTD compressed). New files are synced daily around midnight UTC via GitHub Actions.

What's in Each File

Every row is a single GDELT event. Columns fall into four groups:

~76 columns total. Full schema details in the source repo.

Processing Pipeline

Raw GDELT 15-minute dumps go through:

1. Filter — keep only events involving Saudi Arabia
2. Clean — remove nulls, empty URLs, invalid tone values
3. Deduplicate — one best record per source URL (scored by data completeness)
4. Geo-enrich — match coordinates to nearest city using a 33K+ world cities reference via Haversine distance
5. Scrape — extract article text using Trafilatura → Newspaper4k → Playwright fallback chain (up to 500/day)

Quick Query (DuckDB)

1-- read a single day
2SELECT * FROM 's3://us-west-2.opendata.source.coop/tabaqat/gdelt-sa/country=SA/year=2026/2026_01_01.parquet' LIMIT 10;
3
4-- count events with articles across all days
5SELECT count(*) AS total,
6       count(ArticleTitle) AS with_articles
7FROM 's3://us-west-2.opendata.source.coop/tabaqat/gdelt-sa/country=SA/**/*.parquet';
8
9-- top mentioned cities
10SELECT NearestCity, count(*) AS n
11FROM 's3://us-west-2.opendata.source.coop/tabaqat/gdelt-sa/country=SA/**/*.parquet'
12WHERE NearestCity IS NOT NULL
13GROUP BY 1 ORDER BY 2 DESC LIMIT 10;
1-- read a single day
2SELECT * FROM 's3://us-west-2.opendata.source.coop/tabaqat/gdelt-sa/country=SA/year=2026/2026_01_01.parquet' LIMIT 10;
3
4-- count events with articles across all days
5SELECT count(*) AS total,
6       count(ArticleTitle) AS with_articles
7FROM 's3://us-west-2.opendata.source.coop/tabaqat/gdelt-sa/country=SA/**/*.parquet';
8
9-- top mentioned cities
10SELECT NearestCity, count(*) AS n
11FROM 's3://us-west-2.opendata.source.coop/tabaqat/gdelt-sa/country=SA/**/*.parquet'
12WHERE NearestCity IS NOT NULL
13GROUP BY 1 ORDER BY 2 DESC LIMIT 10;

License & Attribution

Data derived from the GDELT Project. Refer to GDELT's terms of use for redistribution rights. Pipeline source code is MIT licensed.