This GitHub repository contains the codebase for a proof of concept (PoC) implementing the CFR2SBVR (Code of Federal Regulations to SBVR) methodology. The objective is to automate the transformation of financial regulations written in natural language into business rules and vocabulary compliant with the Semantics of Business Vocabulary and Business Rules (SBVR) standard. The repository leverages advanced Natural Language Processing (NLP) techniques, integrating the Financial Industry Business Ontology (FIBO) to address linguistic ambiguities and the Financial Regulation Ontology (FRO) to represent the regulatory content. This automation framework aims to streamline Governance, Risk, and Compliance (GRC) management in financial institutions, which face significant financial losses annually due to non-compliance and ambiguously worded regulations.
The repository includes scripts, pre-trained models, and transformation pipelines that parse the Code of Federal Regulations (CFR), extract relevant regulatory statements, and map them into SBVR-compliant rules and vocabulary in Structured English (SE). The system ensures high precision by using domain-specific ontologies, such as FIBO and FRO, to validate terminology and resolve ambiguities. Additionally, evaluation metrics and experimental results are provided to assess the accuracy and formalism of the automated transformations. By implementing this methodology, the repository contributes to reducing regulatory interpretation errors, enabling efficient compliance processes and supporting financial institutions in meeting stringent regulatory requirements.
For how to use this repository, please refer to the README.