Comparison and Synergy Between Fact-Orientation and Relation Extraction for Domain Model Generation in Regulatory Compliance

摘要

Modern enterprises need to treat regulatory compliance in a holistic and maximally automated manner, given the stakes and complexity involved. The ability to derive the models of regulations in a given domain from natural language texts is vital in such a treatment. Existing approaches automate regulatory rule extraction with a restricted use of domain models counting on the knowledge and efforts of domain experts. We present a semi-automated treatment of regulatory texts by automating in unison, the key steps in fact-orientation and relation extraction. In addition, we utilize the domain models in learning to identify rules from the text. The key benefit of our approach is that it can be applied to any legal text with a considerably reduced burden on domain experts. Early results are encouraging and pave the way for further explorations.