Prof. Dr. Felix Naumann

Thorsten Papenbrock

Research Assistant, PhD Candidate

für Softwaresystemtechnik
Prof.-Dr.-Helmert-Straße 2-3
D-14482 Potsdam
Room: G-3.1.09


Phone: +49 331 5509 294
Email:  thorsten.papenbrock(a)hpi.de
Profiles: Xing
Research: GoogleScholar, DBLP, ResearchGate



Research Interests

Data Profiling:

Solving computationally complex tasks is a challenge and a central activity in data profiling. This involves primarily the discovery of metadata in many gigabyte-sized datasets, which is why algorithms developed for this purpose need to be efficient and robust. Because data profiling offers such a plethora of challenging, yet unsolved tasks, I have chosen it as my primary research area. I am in particular interested in the discovery of data dependencies, such as inclusion dependencies, unique column combinations, functional dependencies, order dependencies, matching dependencies, and many more.

Data Cleansing:

Data is one of the most important assets in any company. Therefore, it is crucial to ensure its quality and reliability. Data cleansing and data profiling are two essential tasks that - if performed correctly and frequently - help to guarantee data fitness. In this area, I am particularly interested in (semi-)automatic duplicate detection methods and normalization techniques as well as their efficient implementation.

Parallel and Distributed Systems:

Due to the complexity of many tasks in IT, a clever algorithm alone is often not able to deliver a solution in time. In these cases, parallel and distributed systems are needed. Especially when facing ever larger datasets, i.e., big data, we need to consider technologies such as map-reduce (e.g. Spark and Flink), actors (e.g. Akka), and GPUs (e.g. CUDA and OpenCL) to implement scalability into our solutions.



  • Database Systems I (2013, 2014, 2015, 2016, 2017)
  • Database Systems II (2013)
  • Data Profiling and Data Cleansing (2014)
  • Information Integration (2015)
  • Data Profiling (2017)
  • Distributed Data Analytics (2017)


  • Advanced Data Profiling (2013)
  • Proseminar Information Systems (2014)

Bachelor Projects:

  • Data Refinery - Scalable Offer Processing with Apache Spark (2015/2016)

Master Projects:

  • Joint Data Profiling - Holistic Discovery of INDs, FDs, and UCCs (2013)
  • Metadata Trawling - Interpreting Data Profiling Results (2014)
  • Approximate Data Profiling - Efficient Discovery of approximate INDs and FDs (2015)
  • Profiling Dynamic Data - Maintaining Matadata under Inserts, Updates, and Deletes (2016)

Master Thesis:

    • Discovering Matching Dependencies (Andrina Mascher, 2013)
    • Discovery of Conditional Unique Column Combination (Jens Ehrlich, 2014)
    • Spinning a Web of Tables through Inclusion Dependencies (Fabian Tschirschnitz, 2014)
    • Multivalued Dependency Detection (Tim Draeger, 2016)

    Online Courses:

    • Datenmanagement mit SQL (openHPI, 2013)


    RDFind: Scalable Conditional Inclusion Dependency Discovery in RDF Datasets

    Kruse, Sebastian; Jentzsch, Anja; Papenbrock, Thorsten; Kaoudi, Zoi; Quiane-Ruiz, Jorge-Arnulfo; Naumann, Felix in Proceedings of the ACM SIGMOD conference (SIGMOD) 2016 .

    Inclusion dependencies (inds) form an important integrity constraint on relational databases, supporting data management tasks, such as join path discovery and query optimization. Conditional inclusion dependencies (cinds), which define including and included data in terms of conditions, allow to transfer these capabilities to rdf data. However, cind discovery is computationally much more complex than ind discovery and the number of cinds even on small rdf datasets is intractable. To cope with both problems, we first introduce the notion of pertinent cinds with an adjustable relevance criterion to filter and rank cinds based on their extent and implications among each other. Second, we present RDFind, a distributed system to efficiently discover all pertinent cinds in rdf data. RDFind employs a lazy pruning strategy to drastically reduce the cind search space. Also, its exhaustive parallelization strategy and robust data structures make it highly scalable. In our experimental evaluation, we show that RDFind is up to 419 times faster than the state-of-the-art, while considering a more general class of cinds. Furthermore, it is capable of processing a very large dataset of billions of triples, which was entirely infeasible before.
    Further Information
    Tags hpi inclusion_dependencies isg profiling rdfind