Extracting Decision Dependencies and Decision Logic from Text Using Deep Learning Techniques

Decision models are increasingly being used in modeling business processes. Hence, extracting decision models automatically from texts would help decision modellers by reducing modeling time and supporting them in their analysis. In this paper, deep learning techniques are investigated to extract decision dependencies and conditional clauses directly from text. By using a large dataset of labeled and tagged sentences and NLP, deep learning techniques (BERT and BI-LSTM-CRF) are trained and tested on the identification of these items. The results show that the performance is sufficiently high to extract decision dependency and logic (semi)-automatically from text which provides a big step towards automatic decision modelling.