Digital Library

Experiments on gpt-3 assisted process model development

• Carlo Simon
• Stefan Haag
• Lara Zakfeld

(2023). ECMS 2023, 37th Proceedings
Edited by: Enrico Vicario, Romeo Bandinelli, Virginia Fani, Michele Mastroianni, European Council for Modelling and Simulation.
DOI: http://doi.org/10.7148/2023
ISSN: 2522-2422 (ONLINE)
ISSN: 2522-2414 (PRINT)
ISSN: 2522-2430 (CD-ROM)
ISBN: 978-3-937436-80-7
ISBN: 978-3-937436-79-1 (CD) Communications of the ECMS Volume 37, Issue 1, June 2023, Florence, Italy June 20th – June 23rd, 2023

https://doi.org/10.7148/2023-0270

Carlo simon, Stefan haag, Lara zakfeld (2023). Experiments on GPT-3 Assisted Process Model Development, ECMS 2023, Proceedings Edited by: Enrico Vicario, Romeo Bandinelli, Virginia Fani, Michele Mastroianni, European Council for Modelling and Simulation. doi:10.7148/2023-0270

Computer assisted process model development from textual descriptions is still an open research question. Advantages of such a technology lie in shorter development times and possibly a more concise interpretation of the narrative input. A solution to this problem necessarily relies on methods from formal modeling and linguistics. In the latter field, the new GPT-3 model is recognized as a breakthrough that outperforms previous technologies whose limitations hindered success of earlier research in this context. But are GPT-3's capabilities to summarize text, detect cause-and-effect, or to classify terms sufficient to succeed? The presented research describes the results of systematic experiments to use GPT-3 to interpret a textual process description and transform it into a formal representation. The different settings demonstrate how to exploit the capabilities of large language models and how to avoid pitfalls. Although the observations made are promising, further work is needed. The outcome of this paper identifies the direction in which this future research should proceed.