The Data Engineering group at the University of Regensburg contributed three presentations to the 30th European Conference on Advances in Databases and Information Systems (ADBIS 2026), held from September 28 to October 1, 2026, in Orléans, France.
On Thursday, October 1, Jennifer Landes presented the research paper Auditing Eye-Tracking Preprocessing Pipelines Using Intrinsic Data Quality Metrics (externer Link, öffnet neues Fenster), co-authored with Meike Klettke. The paper introduces NIMBUS, an approach for evaluating eye-tracking preprocessing pipelines without requiring ground-truth labels. By analyzing intrinsic data-quality metrics across different datasets and preprocessing configurations, the work provides insights into how preprocessing decisions affect signal quality, data retention, and trajectory characteristics.
On Wednesday, September 30, Maximilian Plazotta and Meike Klettke presented their demonstration HARPY: Human-AI Collaboration for Database Error Resolution via LLMs (externer Link, öffnet neues Fenster). HARPY combines large language models with human oversight to identify and resolve database errors. Its interactive interface enables users to review detected anomalies and supports more transparent, human-guided data cleaning.
Also on Wednesday, Meike Klettke delivered an invited lecture entitled Schema Evolution in Relational, NoSQL and Graph Databases as part of the ADBIS Doctoral Consortium School. The lecture explored the challenges of evolving database schemas across different data models and database technologies.
Together, these contributions highlight the group's research activities in data quality, intelligent data cleaning, database evolution, and reliable data preprocessing.
