Document relevance model for Woo requests

Ministry of Health, Welfare and Sport

The relevance model helps lawyers handle Woo requests faster by efficiently dividing documents for a Woo request into 'relevant' and 'not relevant'.

Last change on 19th of June 2025, at 12:25 (CET) | Publication Standard 1.0

• Publication category: Other algorithms
• Impact assessment: DPIA
• Status: Out of use

General information

Theme

• Organisation and business operations
• Law

Begin date

2022-07

End date

2023-12

Contact information

[e-mail]

Responsible use

Goal and impact

The aim is to reduce the time taken to go through the Woo request process by allowing lawyers to make decisions faster. In this way, citizens can get a response to their Woo request sooner. The algorithm pre-selects documents so that lawyers can immediately start working on the most relevant documents. The impact on companies and citizens is minimal because the irrelevant documents are still reviewed at a later date and revealed if necessary.

Considerations

The disadvantage of this algorithm may be that a document relevant to the Woo request may be missed at an early stage. On the other hand, citizens can get faster responses to their Woo requests. Going through the process with the manual method takes much more time than is legally justified, using the relevance model speeds up this process.

Human intervention

The lawyers check whether a document has indeed been given the correct classification (relevant or not relevant).

Risk management

In the end, all documents are reviewed by lawyers, so there is little chance that a relevant document will not be published.

Legal basis

The Open Government Act (Woo) regulates the right to information about everything the government does. It is the successor to the Open Government Act (Wob).

Links to legal bases

• Woo: https://wetten.overheid.nl/BWBR0045754/
• Wob: https://wetten.overheid.nl/BWBR0005252/

Impact assessment

Data Protection Impact Assessment (DPIA)

Operations

Data

Documents (e.g. emails, Office documents, etc.) produced and received by the ministry regarding COVID-19 and reviews by lawyers on these documents.

Technical design

The model is trained on texts of documents reviewed by lawyers. In doing so, the model learns which words do and do not appear in relevant documents. Word weights are used to determine whether a document is relevant. A validation took place on documents labelled by lawyers that were not included in training the model. From this, it could be concluded that the model performs similarly to a lawyer.

External provider

Internally developed