Human-Centred Automated Reasoning for Regulatory Reporting via Knowledge-Driven Computing

• Published in: Trends in Artificial Intelligence Theory and Applications. Artificial Intelligence Practices Vol. 12144; pp. 393 - 406
• Main Authors: Thilakarathne, Dilhan J., Al Haider, Newres, Bosman, Joost
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2020; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Banking; Compliance; Explainability; Finance; Human-centred; Knowledge systems; Reasoning; RegTech; Regulations; Rule-based systems
• ISBN: 303055788X; 9783030557881
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-55789-8_35

The rise in both the complexity and volume of regulations in the regulatory landscape have contributed to an increase in the awareness of the level of automation necessary for becoming fully compliant. Nevertheless, the question of how exactly to become fully compliant by adhering to all necessary laws and regulations remains. This paper presents a human-centred, knowledge-driven approach to regulatory reporting. A given regulation is represented in a controlled natural language form including its metadata and bindings, with the assistance of subject matter experts. This representation of a semi-formal controlled natural language translates into a self-executable formal representation via a context-free grammar. A meta reasoner with the knowledge to execute the given self-executable formal representation while generating regulatory reports including explanations on derived results has been developed. Finally, the proposed approach has been implemented as a prototype and validated in the realm of financial regulation: Money Market Statistical Reporting Regulation.