An Expert-Validated LLM Framework for Transforming Legal Procurement Texts into Actionable Data

Legal and administrative sources typically describe procedural steps that are not recorded in structured data, which makes the application of process-oriented analysis particularly challenging. In this work, we present an approach that uses Large Language Models (LLMs) to extract events and dates from unstructured legal texts. The methodology is applied to a dataset of Italian procurement notices published since 2022 on the official EU platform, Tenders Electronic Daily (TED), demonstrating how LLMs can extract valuable information, such as administrative decisions adopted prior to tender publication. These elements are incorporated into existing event logs, thereby enhancing the quality of process analysis. A sample of the extracted data has been manually reviewed by legal experts to assess the relevance and correctness of the automated detection. The results suggest that this approach can help identify procedural steps hidden in free text, thereby supporting more complete and accurate representations of legal workflows.