Methodology

Pipeline

Every dataset in the catalog goes through the same four steps: fetch a specific, checksummed snapshot of a public dataset; validate it against a strict schema that fails loudly on anything unexpected; normalize the raw fields into stable, documented public columns; publish CSV, JSON and a machine-readable datapackage.json, each with a SHA-256 checksum.

Nothing is published from memory or from a model’s general knowledge. Every number on every page is computed at build time from the files in the catalog—statistics, counts and totals are recomputed each time the data changes, not hand-written.

No invented facts

This site adds no facts that are not in the source data. It does not describe places, amenities or events that aren’t represented by a column in the dataset. Page text is generated from fixed templates filled in with data; nothing about an individual record is written by hand, and no text is generated by AI.

Corrections

If you find an error, email the address on the Contact page with the dataset and what you believe is wrong. Each report is checked against the source data. If the error comes from how this site processed the data, the dataset is corrected and republished as a new version with a changelog entry. If the source data itself is wrong, this site cannot change it, because published rows must match the official source; the report will be answered with that explanation.

Licensing and attribution

Derived datasets are published under CC BY 4.0. Each dataset page and its datapackage.json carry the exact attribution text for every upstream source, including the required modifications list.