The Data Science & Intelligent Systems (DSIS) rled by Prof. Dr. Elena Demidova at the University of Bonn and the Lamarr Institute for Machine Learning and Artificial Intelligence, has achieved a remarkable success at the 22nd International Semantic Web Conference (ISWC 2023). Two DSIS research papers were awarded with the most prestigious prizes of the conference.
The Best Research Track Paper Award was given to the paper "Spatial Link Prediction with Spatial and Semantic Embeddings" by Genivika Mann, Alishiba Dsouza, Ran Yu and Prof. Dr. Elena Demidova.
This work addresses a critical limitation of geographic knowledge graphs — the lack of semantic relationships between geographic entities due to their flat structure. The authors address this challenge with two novel approaches for predicting spatial connectivity in sparsely connected knowledge graphs: supervised spatial connectivity prediction and unsupervised inductive spatial connectivity prediction. These approaches exploit the richness of literal values in geographic knowledge graphs through spatial and semantic embeddings. The evaluation was conducted using the WorldKG knowledge graph developed by the DSIS research group — a comprehensive, large-scale geographic knowledge graph that provides a semantic representation of geographic entities from over 188 countries. Read the full paper here.
The Best Research Track Student Paper Award was given to the paper "Iterative Geographic Entity Alignment with Cross-Attention" by Alishiba Dsouza, Ran Yu, Moritz Windoffer and Prof. Dr. Elena Demidova.
This paper presents Iterative Geographic Entity Alignment (IGEA), a novel cross-attention based iterative alignment approach that addresses the challenges of aligning schemas and entities of collaboratively created geographic data. IGEA overcomes differences in entity representations, sparse linkages and high schema heterogeneity. Experiments with real-world datasets from different countries have shown that IGEA improves entity alignment performance by up to 18 percentage points. The iterative method also improves entity and tag-to-class alignment performance by 7 and 8 percentage points in F1 score, respectively. To the full publication.
We hope that the new algorithms and research results presented in these publications as well as the WorldKG knowledge graph released by the DSIS research group will open up new research opportunities in the field of semantic spatio-temporal data analysis.
Über die DSIS-Forschungsgruppe:
Die Forschungsgruppe Data Science & Intelligent Systems (DSIS), unter der Leitung von Prof. Dr. Elena Demidova an der Universität Bonn und dem Lamarr-Institut für Maschinelles Lernen und Künstliche Intelligenz, entwickelt innovative maschinenlern- und wissensbasierte, interaktive, transparente und erklärbare Data Science-Methoden für große, heterogene Datensätze.
Zu den Forschungsinteressen des DSIS gehören raum-zeitliche und mehrsprachige Daten, Open Data, das Web und das Semantic Web. Zu den Anwendungsbereichen gehören Mobilität, intelligente Städte, Versorgungsketten und das Gesundheitswesen.
Für weitere Informationen besuchen Sie bitte die Homepage.
Über die ISWC 2023:
Die 22. Internationale Semantic Web Konferenz (ISWC 2023) ist das wichtigste internationale Forum für die Semantic Web Community. Die ISWC 2023 bringt Forscher, Praktiker und Branchenexperten zusammen, um die Zukunft der semantischen Technologien zu diskutieren, voranzutreiben und zu gestalten. Weitere Informationen finden Sie auf der Website der Konferenz.
Danksagung
Die Arbeit an diesen Veröffentlichungen wurde mit Mitteln der DFG, Deutsche Forschungsgemeinschaft ("WorldKG", 424985896), des Bundesministeriums für Wirtschaft und Klimaschutz (BMWK), Deutschland ("ATTENTION!", 01MJ22012C), und des DAAD/BMBF, Deutschland ("KOALA", 57600865) gefördert.