Extracting Decision Model and Notation models from text using deep learning techniques

Companies and organizations often use manuals and guidelines to communicate and execute operational decisions. Decision Model and Notation (DMN) models can be used to model and automate these decisions. Modeling a decision from a textual source, however, is a time intensive and complex activity hence a need for shorter modeling times. This paper studies how NLP deep learning techniques can extract decision models from text faster. In this paper, we study and evaluate an automatic sentence classifier and a decision dependency extractor using NLP deep learning models (BERT and Bi-LSTM-CRF). A large labeled and tagged dataset was collected from real use cases to train these models. We conclude that BERT can be used for the (semi)-automatic extraction of decision models from text.