Detecting Constraints and Their Relations from Regulatory Documents Using NLP Techniques

摘要

Extracting constraints and process models from natural language text is an ongoing challenge. While the focus of current research is merely on the extraction itself, this paper presents a three step approach to group constraints as well as to detect and display relations between constraints in order to ease their implementation. For this, the approach uses NLP techniques to extract sentences containing constraints, group them by, e.g., stakeholders or topics, and detect redundant, subsuming, and conflicting pairs of constraints. These relations are displayed using network maps. The approach is prototypically implemented and evaluated based on regulatory documents from the financial sector as well as expert interviews.