Legal process mining for the analysis of public procurement workflows

Public procurement procedures are often characterised by complex workflows, which can pose challenges in terms of transparency, compliance, and operational efficiency. Automated process analysis is particularly important for understanding the complexity of temporal events, as it facilitates the understanding of the contents of procurement procedures by citizens and helps to increase trust in public authorities. This study explores process mining techniques applied to legal data with the aim of understanding how these procedures take place. By reconstructing event logs from the official dataset, we analyse the actual sequence of activities involved in the tendering process. First, we demonstrate how Large Language Models can be leveraged to extract relevant information from legal texts, thus enriching the event log with additional case-specific details, including dates and financial data. Second, we introduce a prediction task that estimates the overall duration of a tender process based on its initial activities. Through this combined approach, we investigate whether integrating textual insights enhances the accuracy of predictive models. The analysis is grounded in a real-world dataset. It offers evidence of how the integration of Natural Language Processing and process-oriented methods can support informed decision-making.