Leveraging Large Language Models (LLMs) for Process Mining (Technical Report)

• DOI: 10.48550/arxiv.2307.12701
• Publication Date: 2023-01-01
• Authors: Alessandro Berti; Mahnaz Sadat Qafari
• Keywords: Process mining; Abstraction; Process modeling; Business process discovery

This technical report describes the intersection of process mining and large language models (LLMs), specifically focusing on abstraction traditional object-centric artifacts into textual format. We introduce explore various prompting strategies: direct answering, where model directly addresses user queries; multi-prompt which allows to incrementally build knowledge obtained through a series prompts; generation database queries, facilitating validation hypotheses against original event log. Our assessment considers two models, GPT-4 Google's Bard, under contextual scenarios across all strategies. Results indicate that these exhibit robust understanding key abstractions, with notable proficiency in interpreting both declarative procedural models. In addition, we find demonstrate strong performance setting, could significantly propel advancement discipline. Additionally, display noteworthy capacity evaluate concepts fairness mining. opens door more rapid efficient assessments logs, has significant implications for field. The integration applications may open new avenues exploration, innovation, insight