Prof. Dr. Felix Naumann
  • GermEval 2019
    Our paper hpiDEDIS at GermEval 2019: Offensive Language Identification using a German BERT model has been accepted at the GermEval Workshop at KONVENS 2019. It is a joint research project of the Hasso Plattner Institute at the University of Potsdam (HPI) and the Research Group on Deliberative Discussions in the Social Web at the Heinrich Heine University Düsseldorf (DEDIS). We published a preprint here and our source code on GitHub here.
  • Data Repeatability in Web Science
    Our article Measuring and Facilitating Data Repeatability in Web Science (Link to pre-print) has been accepted for publication in the Datenbank-Spektrum Journal. We publish source code corresponding to this article here.
  • Toxic Comment Classification
    We participated in the Toxic Comment Classification Challenge (Link), which was a Kaggle challenge with the goal to identify and classify toxic online comments. In collaboration with our colleagues from the DATEXIS group at Beuth Hochschule für Technik Berlin, we finished in the top 2% of the leaderboard and achieved 54th place out of 4551 teams.
  • Aggression Identification
    We participated in the First Shared Task on Aggression Identification (Link), which is part of the First Workshop on Trolling, Aggression and Cyberbullying at the 27th International Conference of Computational Linguistics (COLING 2018). Our team achieved 2nd place out of 30 teams at the task of classifying social media posts as ‘Overtly Aggressive’, ‘Covertly Aggressive’, or ‘Non-aggressive’ on an unseen test dataset. Our implementation and our augmented dataset is published here. The paper is uploaded here.
  • Semi-Automated Comment Moderation 
    You can find code that accompanies our paper Delete or not Delete? Semi-Automatic Comment Moderation for the Newsroom here. The paper itself can be found here.