About this translation

CBS publishes its microdata documentation in Dutch. Researchers still work with Dutch variable names, so a literal translation would not help: this site uses English as a lens on the Dutch.

Who makes this

DIEP · University of Amsterdam

A prototype developed by the Dutch Institute for Emergent Phenomena (DIEP) within the University of Amsterdam Research Priority Area Emergent Phenomena in Society: Polarization, Segregation and Inequality (RPA PSI). This prototype is part of the RPA PSI project at the University of Amsterdam and acknowledges UvA support for the research priority area.

CBS Guru started as a retrieval prototype for finding CBS variables from a project description. This site adds an English reading layer on top of the Dutch documentation.

How a page is made

1. Extract. Each CBS PDF is parsed into datasets, chapters, variables, definitions, notes, code lists and version history.
2. Protect the identifiers. Variable names, dataset codes and file names are replaced by placeholders before translation and restored afterwards, so they cannot be changed.
3. Translate. An AI model translates each unique block once, keeping Dutch terms in brackets at first use.
4. Check automatically. Every translation must keep all placeholders, all numbers and all capitalised identifiers of the source. Failures are listed and fixed.
5. Link back. Each variable shows its PDF page, and each dataset links to the original PDF.

What the status badges mean

AI-translated The text was produced by an AI model.

Checked automatically Identifiers, numbers and placeholders match the source. This says nothing about the quality of the prose.

Reviewed A named Dutch speaker has read the page against the original. No page has this status yet.

Reporting a problem and reviewing

Click “report” on any variable. It copies the dataset and variable, ready to paste into a new GitHub issue with your correction. Reviewers can switch on “Show Dutch original” and read the English and Dutch side by side. Corrections go into the translation files and the site is rebuilt; the status changes to Reviewed only when a reviewer signs off.

Limits of this pilot

• 411 of 687 CBS documents. The older document layout is not handled yet.
• Dates, definitions and thresholds can change; the CBS PDF is always the reference.
• This is not official CBS material and is not affiliated with CBS.
• Permission to republish CBS documentation in translation still has to be confirmed before a public launch.

Built

2026-11-12