Rule-based Information Extraction for KIDs: The PRIIPs Information Extraction Framework (PRIIEF) Supports Compliance, Oversight, and Education

Sammanfattning

This thesis addresses the challenge of improving regulatory compliance in financial product disclosures by applying rule-based information extraction (IE) to Key Information Documents (KIDs) and related marketing materials under the PRIIPs (Packaged Retail and Insurance-based Investment Products) Regulation. Two key research gaps are identified: the dismissal of rulebased IE for tabular data and the limited understanding of the broader range of compliance errors found in KIDs and associated marketing materials. To address these, a Design Science Research approach was used to develop the PRIIPs Information Extraction Framework (PRIIEF), which informed the design of a rule-based prototype. Tested on 463 Swedish KIDs and associated marketing materials, the prototype achieved over 99% precision, recall, and F-measure for both free text and tabular data – outperforming AI (Artificial Intelligence)-based solutions applied in prior research, and uncovered several compliance violations – such as the absence of any indication to the KID in the analyzed marketing materials, as required under Article 9 of the PRIIPs Regulation. The results show that rule-based IE, when guided by a structured framework aligned with PRIIPs regulatory requirements and the KID template, enables systematic, transparent, and accurate supervision while supporting manual regulatory assessments.