Methodology
How the atlas is built
Where our data comes from, how we decide what belongs, and how we keep it honest.
Sources
- OpenStreetMap (© OpenStreetMap contributors, ODbL 1.0), queried through the Overpass API and, at low volume, the Nominatim search service — always within their published usage policies, with an identifying client and rate limits.
- Restaurant owners through verified claims, and the public through suggestions and corrections.
- Administrator imports, where each row must state its source and the licence or permission under which we may use it.
We do not scrape Google Maps or review sites, and we never bypass logins, CAPTCHAs, robots rules or API restrictions. Every fact we store keeps a record of which source supplied it and when; each profile shows its sources.
Search terms
Discovery looks for structured cuisine tags (such as cuisine=kerala, malabar, thalassery, travancore, syrian_christian, sadya) and for names that describe Kerala food — for example “Kerala Kitchen”, “Thalassery Restaurant”, “thattukada”, “sadya”, “appam”, “karimeen”.
Deciding what belongs
- Structured cuisine data that names a Kerala cuisine is the strongest automatic signal.
- The business’s own name describing its food (“Kerala Cuisine”, “Malabar Restaurant”, “Thattukada”) is a medium signal.
- An ambiguous name (a bare “Malabar”, which is also a place and a brand) is held for human review and not published.
- An owner’s name, ethnicity or presumed background is never used.
Automatically accepted listings are published as Unverified with a confidence score and the reason they were included, so you can judge for yourself.
Duplicates and branches
New records are compared with existing listings by name similarity and distance. Near-identical records within about 150 m are merged; close but uncertain matches are queued for an editor. Separate locations of the same restaurant are kept as branches.
Missing information
If no permitted source provides a field — phone, opening hours, price, menu — it stays blank and a verification task is created. We would rather show less than show something invented.
Freshness
Sources are re-checked on a schedule, and every profile shows when it was last checked by a person. Opening hours in particular change often; please confirm before you travel.
Awards
Listing data and awards are separate. See how awards are decided.