Investigating the Use of Natural Language Processing (Nlp) Techniques in Automating the Extraction of Regulatory Requirements from Unstructured Data Sources

Abstract

In this paper, the focus is placed on proposing an approach that deals with the automation of the identification of regulatory requirements from text documents using NLP techniques. Hence, it provides a way of enhancing the identification of regulatory requirements from manuals, policy statements, and documents; these use NLP. The study focuses on methods of collecting and cleansing data, the procedure of developing and forming NLP models, as well as the process of assessing and enhancing the formed models. It is evident from the study that the regulatory requirements can be extracted with moderate efficiency and accuracy and with advanced transformer models offering higher results as compared to the traditional machine learning algorithms. It has also recognized the challenges faced when working on saturated regulation text and, as stated in the study, the decreasing of compliance processes through NLP. The paper concludes the best practices for future research that are designed to strengthen the contextual understanding and optimization of the NLP models in the conditions of emerging regulations.