Extracting Decision Models from Textual Descriptions of Processes

Decision models are strategic for formalizing how data influences the main decisions in a organization. Due to its importance, standard notations like DMN have appeared in recent years, to serve as a central resource for synchronizing the people and systems with respect to decisions. However, the modeling of DMN specifications can be tedious and error-prone, hampering its adoption in practice. This paper presents a technique to automatically obtain complete DMN models from textual descriptions. The technique, grounded in natural language processing combined with tailored syntactic patterns, allows to extract both the decision requirements and the decision logic described in a text. Our experimental evaluation shows promising results, even for the quite small pattern set used.