| Acronym: OptiGov |
| Main Objective: Public administration processes are inherently complex due to numerous factors, with diverse executions often diverging from the ‘happy paths’ envisioned by legislators and the high number of guidelines and requirements with which they should comply. This project aims to harness the power of Data Science and Artificial Intelligence (AI), particularly Process Mining (PM) and Large Language Model (LLM) based techniques, to optimize and manage this complexity. Our project goals are multifaceted. At the core, we aim to introduce a general-purpose method for log-based compliance, by aligning specification documents, enriched with contextual information about the executions of the involved processes, and the meta-level guidelines dictating how these processes should be. This novel approach will draw inspiration from the symbolic techniques used in data-aware declarative process mining, where the alignment is with ‘normative processes’ expressed via declarative rules. A challenge that we should address is to handle specification documents and guidelines that provide meta-level information about business requirements that the process should satisfy, rather than executions and process models. Our method will be significantly enhanced with LLM techniques to preprocess the textual documents and return symbolic representations that enable effective alignment computations. Firstly, we aim to develop a structured symbolic formalism to express the constraints and normative rules that guide the specifications of PA processes. We will also record events and activities of the underlying PA processes in a data-aware event log to capture the actual executions, providing a context-rich basis for analysis. Secondly, we will utilize LLMs to extract relevant information from legal guidelines and the specification documents and to transform it into this structured symbolic formalism. We will enrich the specification documents with the information mined from the event logs. Third, we intend to develop a new method for computing alignments, which will systematically identify and quantify the core inconsistencies between the enriched specification documents (augmented with business knowledge from event logs) and the legal guidelines. This method will provide a quantitative score representing the degree of alignment, offering a precise measure of how closely the specifications adhere to the legal requirements. Finally, we will apply these techniques to the “ICT governance process in PA entities”, assessing the alignment of technical specifications (in documents, e.g., PDFs) of enterprise architectural scenarios (submitted by any PA entity with architectural principles) with general ICT guidelines for these entities. This pilot will create solid foundations to ensure compliance with Portuguese Decree-Law 107/2012 (“Parecer Prévio TIC”). The ‘Agência para Modernização Administrativa’ (AMA) and the ‘Instituto de Gestão Financeira e Equipamentos da Justiça’ (IGFEJ) agreed to collaborate in this project. The benefits of this project for the Portuguese PAs include enhancing compliance by accurately aligning specification documents with legal guidelines, thereby sparing time and money in the evaluation process, optimizing process workflows, increasing efficiency, providing actionable insights for decision-making, and improving service delivery to citizens. This project aspires to foster a more efficient governmental process that benefits all stakeholders. |
| Reference: 2024.07385.IACDC |
| Funding: FCT |
| Start Date: 01-02-2025 |
| End Date: 31-01-2026 |
| Team: Chrysoula Zerva, Alessandro Gianola |
| Groups: Pattern and Image Analysis – Lx |
| Partners: INESC-ID |
| Local Coordinator: Chrysoula Zerva |
|
|
|
|
|
Associated Publications
|
