Anonymising documents Woo requests

Municipality of Nijmegen

According to the Open Government Act (Woo), certain parts of documents must be made anonymous (anonymised). In that case, they no longer contain any data that could identify individuals. To this end, the municipality of Nijmegen uses a computer programme with an algorithm that recognises (personal) data and other confidential information.

Last change on 20th of February 2025, at 15:33 (CET) | Publication Standard 1.0

• Publication category: Other algorithms
• Impact assessment: DPIA
• Status: In use

General information

Theme

• Organisation and business operations
• Law

Begin date

2023-07

Contact information

[e-mail]

Link to publication website

https://nijmegen.nl/diensten/wet-open-overheid

Responsible use

Goal and impact

We use the computer anonymisation programme to ensure that we can disclose as much information as possible (transparency). And at the same time to protect individuals, companies and organisations that may be named in the documents. The computer programme helps us do anonymisation as quickly and easily as possible.

The computer programme's algorithm poses little risk. This applies to individuals (residents and employees of organisations using the programme), companies and organisations. The algorithm specifically searches for (personal) data and hides or reports them, whatever else is in the documents. The computer programme makes a proposal for anonymising a piece of text. An employee with knowledge of the subject assesses the proposal: the programme therefore does not make automatic decisions. Furthermore, an employee within the programme can hide (black mark) text that cannot be disclosed for other reasons. For example, a piece of text containing strategic information for the protection of the municipality or organisations we work with. The document always states the reason for hiding text.

Considerations

Without the use of the computer programme, anonymising documents takes much more time. Using the programme allows us to disclose information more quickly and easily. There is also less chance of human error due to automatic anonymisation. This reduces the chance of a data leak and ensures better protection of individuals' data.

If a subject does not involve many documents to be disclosed, we anonymise them ourselves. As a result, we use the algorithm only when necessary.

Human intervention

The algorithm makes proposals that are always checked by hand. So there is a human check on the algorithm. We can also specify which words should or should not be anonymised anyway.

Risk management

The risks are anonymising too much or too little. To ensure this does not happen, a staff member checks the algorithm's proposals.

Legal basis

The Open Government Act (Woo), which regulates disclosure of documents. The General Data Regulation (AVG) applies to the meaning of personal data.

Links to legal bases

• Wet open overheid: https://wetten.overheid.nl/BWBR0045754/2024-08-01/0
• Algemene verordening Persoonsgegevens: https://eur-lex.europa.eu/legal-content/NL/TXT/HTML/?uri=CELEX:02016R0679-20160504

Link to Processing Index

https://www.nijmegen.nl/verwerkersnijmegen.php

Impact assessment

Data Protection Impact Assessment (DPIA)

Operations

Data

This depends on the document being anonymised. Examples include personal data such as e-mail addresses, phone numbers, bank account numbers, address details and signatures. And based on the Open Government Act (Woo), it can also include data beyond personal data. These exceptions are set out in the Woo.

Technical design

Deep learning models that determine which information is considered privacy-sensitive. The models do this by assessing how pages look (visual inspection) and by scanning text.

External provider

eData B.V.