Information Extraction and Classification of Documents for Income Tax (IECD_IH)

Dutch Tax and Customs Administration

This page provides information about the algorithm 'IECD_IH' - Information Extraction and Classification of Documents for Income Tax. This algorithm reads incoming letters to categorise them to the appropriate department within the Inland Revenue.

Last change on 12th of September 2025, at 14:43 (CET) | Publication Standard 1.0

• Publication category: Impactful algorithms
• Impact assessment: Field not filled in.
• Status: In use

General information

Theme

Public finance

Begin date

Field not filled in.

Contact information

[e-mail]

Link to publication website

https://over-ons.belastingdienst.nl/onderwerpen/omgaan-met-gegevens/algoritmeregister/

Link to source registration

https://over-ons.belastingdienst.nl/onderwerpen/omgaan-met-gegevens/algoritmeregister/informatie-extractie-en-classificatie-van-documenten-voor-inkomensheffing-iecd_ih/

Responsible use

Goal and impact

Various objection and request letters arrive by mail. This algorithm categorises these letters so that they go directly to the right department for further processing.

If the algorithm cannot categorise a letter, an employee manually reads the letter and ensures that it is still submitted to the correct department. Letters that may have been incorrectly categorised are also corrected manually by an employee

Considerations

Given the large numbers of incoming objection and request letters, we use an algorithm to categorise them automatically. This method is more consistent and ensures that letters reach the right department for further processing more quickly.

Another key advantage of this way of working is increased efficiency: because the algorithm takes over much of the categorisation work, less human effort is required. Objections and requests are picked up faster, significantly reducing the processing time.

Human intervention

Human intervention in the Inland Revenue context implies that a competent and knowledgeable employee plays a substantial role in the decision-making process.

The algorithm IECD IH does not perform substantive decisions. It only assists in assigning letters to the correct department. Assignment requires human intervention in some cases. Whereby the final assessment of an objection or request always lies with a Tax Department staff member. Employees make the decisions and ensure that each file is handled carefully and correctly.

This algorithm does not make decisions.

Risk management

The General Administrative Law Act (Awb) requires the government's actions to be transparent and lawful. The Tax and Customs Administration observes the general principles of good governance when applying and developing algorithms.

Conditions, a quality framework, have been drawn up by the Tax and Customs Administration for the development of algorithms. This contains rules and agreements that are followed during algorithm development. The conditions of the National Audit Authority are leading in this respect. At set moments, the Tax and Customs Administration checks whether the algorithm still meets the quality requirements.

The algorithm uses data collected under various tax laws. As required by the AVG, no more data is used than necessary.

Legal basis

1. General State Tax Act:
2. General Administrative Law Act:
3. General Data Protection Regulation:
4. General Data Protection Regulation Implementation Act:
5. Income Tax Act 2001:
6. Archives Act 1995:

Links to legal bases

• General State Tax Act:: https://wetten.overheid.nl/BWBR0002320/
• General Administrative Law Act:: https://wetten.overheid.nl/BWBR0005537/
• General Data Protection Regulation:: https://eur-lex.europa.eu/legal-content/NL/TXT/HTML/?uri=CELEX:32016R0679
• General Data Protection Regulation Implementation Act:: https://wetten.overheid.nl/BWBR0040940/
• Income Tax Act 2001:: https://wetten.overheid.nl/BWBR0011353/
• Archives Act 1995:: https://wetten.overheid.nl/BWBR0007376/

Elaboration on impact assessments

• Privacy and AVG

The use of data should be assessed against the General Data Protection Regulation (AVG). Reviewing personal data reveals any privacy risks and allows appropriate measures to be taken.

The AVG prescribes that no more data should be used than necessary. This is called data minimisation. The Tax Administration regularly examines whether the data used are still necessary and may therefore be used.

• Use of special personal data

No special personal data are used in this algorithm.

• Equality and non-discrimination

The algorithm is assessed in line with applicable non-discrimination principles for direct and indirect discrimination. Processing as little personal data as possible reduces the risk of direct discrimination. Employees involved in developing and managing the algorithms receive training on data protection and bias.

Operations

Data

1. The contents of the letter
2. Case information
3. VIP indication of person

Links to data sources

• The contents of the letter: De brief
• Case information: Logistieke Werkstroom Besturing (LWB/zaakadministratie belastingdienst) en Teradata (Datawarehouse architectuur waarin onder andere zaaknummers worden vastgelegd)
• VIP indication of person: Interne gegevens Belastingdienst

Technical design

It is a learning algorithm: the algorithm has been trained by humans based on past data to discover connections. The algorithm is periodically re-trained by humans.

The algorithm's outcome is which process a request or objection belongs to. When the algorithm cannot determine this, the letter is manually categorised by an employee.

The algorithm is not self-learning. This means that the algorithm does not develop itself while using it.

External provider

The algorithm was developed by staff at the Inland Revenue and is also maintained internally.