Cem Okulmus
Postdoctoral Researcher at Paderborn University
I currently work as a PostDoc in the Knowledge Representation group at Paderborn University led by Anni-Yasmin Turhan. My recent work has focused on exploring learnability of constraints over graph-structured data in the context of RDF and SHACL.
Previously I was a postdoctoral fellow at Umeå University where I worked with (then) WASP guest professor Diego Calvanese on the topic of temporal ontology-based data-access (OBDA).
I completed my PhD at TU Wien, where my thesis focused on exploring the practical computation of structural decomposition methods for relational queries.
Topics I am interested in include
- Learnability/Separability in various languages over graph-structured data
- Temporal Ontology-Based Data Access
- Ontology-Mediated Query Answering for Graph Queries
- Shape Constraint Languages
- Structural Width Measures over Hypergraphs
CV
Publications
2026
Rewriting Ontology-Mediated Property Graph Queries into GQL Accepted
This is our latest published work on the topic of bringing Ontology-Based Data-Access to the world of graph data bases. Compared to our previous paper (see ESWC 2025), we consider a much richer set of navigational queries as the input of rewriting, and in the output language we fully exploit the expressive power of GQL, the recently standardised graph query language.
sheval: An RDF data shapes evaluation tool and test-suite for recursive shapes Accepted
Shape Learning refers to the process of generating parts of a SHACL schema (the so-called shapes), in this case from labelled examples. The goal is to find a fitting shape, in the sense that it includes all positively-labelled examples and none of the negatively-labelled ones.
In this work, we provide the first algorithm for learning such fitting shapes for a given fragment of SHACL, computing not only arbitrary fittings, but also the most specific fittings.
This is the start of a larger project, where the ultimate aim is to have systems that can automatically generate entire SHACL schemas, with formal guarantees on its correctness w.r.t. the examples the shapes were learned from.
Querying Interval-Based Temporal Data with SPARQL
We present a syntactic extension of SPARQL—built ontop of SPARQL 1.2—that has
- interval temporal variables as first-class objects,
- allows simple joining over these temporal variables, and
- can easily describe user-defined predicate over temporal variables.
This language will serve as the front-end language of a temporal extension of the Ontology-Based Data-Access system ONTOP, which is part of our on-going work.
Common Foundations for Recursive Shape Languages
Reaching for the stars in concept learning
Selective Use of Yannakakis’ Algorithm for Consistent Performance Gains
2025
Query Rewriting for Nested Navigational Queries over Property Graphs
Soft and Constrained Hypertree Width
Towards Practicable Algorithms for Rewriting Graph Queries Beyond DL-Lite
Common Foundations for SHACL, ShEx, and PG-Schema
2024
SHACL Validation under the Well-founded Semantics
One Language to Rule Them All: Behavioural Querying of Process Data Using SQL
Expressive Power and Complexity Results for SIGNAL, an Industry-scale Process Query Language
Towards Practicable Algorithms for Rewriting Graph Queries Beyond DL-Lite (Extended Abstract)
Fast Parallel Hypertree Decompositions in Logarithmic Recursion Depth (Journal article)
2023
Rewriting Ontology-Mediated Navigational Queries into Cypher
On the Way to Temporal OBDA Systems (short paper)
Reaching Back to Move Forward: Using Old Ideas to Achieve a New Level of Query Optimization (short paper)
Structure-Guided Query Evaluation: Towards Bridging the Gap from Theory to Practice.
Incremental Updates of Generalized Hypertree Decompositions
2022
Parallel Computation of Structural Decompositions
Fast Parallel Hypertree Decompositions in Logarithmic Recursion Depth.
Fast and parallel decomposition of constraint satisfaction problems (Journal article)
2020
Fast and Parallel Decomposition of Constraint Satisfaction Problems.
The Hypertrac Project: Recent progress and future research directions on hypergraph decompositions
2019
Parallel Computation of Generalized Hypertree Decompositions (short paper)
Teaching
University Assistant at DBAI group at Faculty of Informatics, TU Wien.
Helped with organising the following courses:- Database Systems
- Winter Terms 2018 - 2022
- Semi-structured Data
- Summer Terms 2019 - 2022
- Advanced Database Systems
- Summer Terms 2020 - 2022
- Database Systems
Teaching Duties as part of the Knowledge Representation group, Paderborn University.
Helped with organising the following courses:- Programming 2
- Summer Term 2025
- Modelling
- Winter Term 2025
- Proseminar: Gems of Theoretical Computer Science
- Project Group: Ontology-mediated Validation of Knowledge Graphs
- Programming 2
Thesis Supervision
- "Multimedia Processing: Real-Time Colour Grading with JIT using the MLT Framework", Pina Kolling, Master Thesis, Umeå University, Supervised by Cem Okulmus and Urban Söderberg, 2024
- "Comparative Study of Solution-Conflict and Query-Conflict Hypergraphs under Tree Decomposition", Chen Cheng, Bachelor Thesis, Paderborn University, Supervised by Yasir Mahmood and Cem Okulmus, 2026
- "A Dynamic Programming-Based Implementation for Consistent Query Answering", Wenghao Bi, Bachelor Thesis, Paderborn University, Supervised by Yasir Mahmood and Cem Okulmus, 2026
Scientific Engagement
Member of the Programme Committee for the following conferences
- JELIA 2023 — 18th edition of the European Conference on Logics in Artificial Intelligence
- KR 2024 — 21st International Conference on Principles of Knowledge Representation and Reasoning
- JELIA 2025 — 19th edition of the European Conference on Logics in Artificial Intelligence
- ECAI 2025 — 28th European Conference on Artificial Intelligence
- KR 2026 — 23rd International Conference on Principles of Knowledge Representation and Reasoning
- CIKM 2026 — 35th International ACM Conference on Information and Knowledge Management
Member of the Programme Committee for the following workshops
Etc
- Presentation Of Interactive Normalform Transformations (point)
Tool I developed as part of teaching a database course to come up with functional dependencies in various normal forms. The source code can be found here. - shaWell
A SHACL validator that I wrote which supports the 1.0 SHACL Core standard, and has support for recursive shapes. It follows the well-founded semantics for SHACL, as proposed in our paper.