arXiv subjects

Iris Beerepoot

Publications and source records attributed to Iris Beerepoot.

5 recordsLinked to original sources

arXiv2026

Shadows of Transparency: Signaling Transparency-Impeding Behavior Using Public Data

Freedom of Information (FOI) laws aim to increase government transparency, yet existing assessments focus mainly on legal compliance and procedural outcomes, leaving organizational behavior underexamined. As FOI processes are increasingly mediated through digital information systems, public response data offer traces of how organizations handle requests and disclosures in practice. This paper develops and evaluates a pattern-based approach that uses such data to signal transparency-impeding behavior across government agencies and over time. Drawing on prior literature, we identify nine recurring behavioral patterns that undermine transparency. Using a dataset of 15,277 Dutch FOI dossiers comprising 139,449 documents, we operationalize and evaluate measurable indicators for six of these patterns, demonstrating that they enable systematic comparison of disclosure practices across agencies and over time. Expert interviews with researchers, journalists, and civil servants confirm the interpretability and practical usefulness of the indicators for investigative prioritization, comparative research, and transparency oversight.

cs.CY↗

arXiv2025

Integrating Wearable Data into Process Mining: Event, Case and Activity Enrichment

In this short paper, we explore the enrichment of event logs with data from wearable devices. We discuss three approaches: (1) treating wearable data as event attributes, linking them directly to individual events, (2) treating wearable data as case attributes, using aggregated day-level scores, and (3) introducing new events derived from wearable data, such as sleep episodes or physical activities. To illustrate these approaches, we use real-world data from one person, matching health data from a smartwatch with events extracted from a digital calendar application. Finally, we discuss the technical and conceptual challenges involved in integrating wearable data into process mining for personal productivity and well-being.

cs.DB↗

arXiv2025

ExOAR: Expert-Guided Object and Activity Recognition from Textual Data

Object-centric process mining requires structured data, but extracting it from unstructured text remains a challenge. We introduce ExOAR (Expert-Guided Object and Activity Recognition), an interactive method that combines large language models (LLMs) with human verification to identify objects and activities from textual data. ExOAR guides users through consecutive stages in which an LLM generates candidate object types, activities, and object instances based on contextual input, such as a user's profession, and textual data. Users review and refine these suggestions before proceeding to the next stage. Implemented as a practical tool, ExOAR is initially validated through a demonstration and then evaluated with real-world Active Window Tracking data from five users. Our results show that ExOAR can effectively bridge the gap between unstructured textual data and the structured log with clear semantics needed for object-centric process analysis, while it maintains flexibility and human oversight.

cs.DB↗

arXiv2025

TAPAS: A Pattern-Based Approach to Assessing Government Transparency

Government transparency, widely recognized as a cornerstone of open government, depends on robust information management practices. Yet effective assessment of information management remains challenging, as existing methods fail to consider the actual working behavior of civil servants and are resource-intensive. Using a design science research approach, we present the Transparency Anti-Pattern Assessment System (TAPAS) -- a novel, data-driven methodology designed to evaluate government transparency through the identification of behavioral patterns that impede transparency. We demonstrate TAPAS's real-world applicability at a Dutch ministry, analyzing their electronic document management system data from the past two decades. We identify eight transparency anti-patterns grouped into four categories: Incomplete Documentation, Limited Accessibility, Unclear Information, and Delayed Documentation. We show that TAPAS enables continuous monitoring and provides actionable insights without requiring significant resource investments.

cs.CY↗

arXiv2023

Turning Logs into Lumber: Preprocessing Tasks in Process Mining

Event logs are invaluable for conducting process mining projects, offering insights into process improvement and data-driven decision-making. However, data quality issues affect the correctness and trustworthiness of these insights, making preprocessing tasks a necessity. Despite the recognized importance, the execution of preprocessing tasks remains ad-hoc, lacking support. This paper presents a systematic literature review that establishes a comprehensive repository of preprocessing tasks and their usage in case studies. We identify six high-level and 20 low-level preprocessing tasks in case studies. Log filtering, transformation, and abstraction are commonly used, while log enriching, integration, and reduction are less frequent. These results can be considered a first step in contributing to more structured, transparent event log preprocessing, enhancing process mining reliability.

cs.DB↗