Study certificates - Automatic recognition for applying for child benefit

This information was sourced from the European Commission, Joint Research Centre (2026): PSTW: Public Sector Tech Watch latest dataset of selected cases. [Dataset] doi: 10.2905/JRC.J1MWC9R PID: http://data.europa.eu/89h/e8e7bddd-8510-4936-9fa6-7e1b399cbd92. Licensed under Creative Commons Attribution 4.0 (CC BY 4.0).

To apply for child benefit, the current certificate of study of the child in question must be submitted twice a year. This results in a high volume of documents to be checked by hand, especially since the German universities could not agree on a uniform form. The client uploads the certificate of study to an appropriate website. The text of the certificate of study is extracted with OCR (Optical Character Recognition) and transferred to the system for classification. Four issues are checked: Is it a certificate of study (yes/no) and with what probability (score)? Is it the right child? Is it the correct validity period/semester? Is it a known German university?

The result of the classification is displayed to the employees in the specialist procedure of the family benefits office. At the same time, the result of the check procedure "Studienbescheinigung ja /nein" (certificate of study yes /no) is indicated. This must then be actively accepted by the employee.

The training data currently consists of 60,000 study certificates that are semi-automatically annotated and then used for training. Using a specially developed algorithm, the personal data is anonymised before classification and can therefore also be used as training data. The classification of the study certificate is carried out by: Naive Base/Support Vector Machine / Logistic Regression, NLP (Natural Language Processing), and Named Entity Recognition is used for data extraction.

Budget: Not disclosed