The jury of the Fakultätentag Informatik for the best computer science thesis awards the prize in 2021 to Marek Herde for his master thesis "Estimating Annotation Performances for Probabilistic Active Learning with Multiple Annotators." The award ceremony will take place on Computer Science Day during the annual conference of the Gesellschaft für Informatik in Berlin, probably on September 30, 2021.

Mr. Herde proposed "Multi-annotator Probabilistic Active Learning" (MaPAL) in his master's thesis as a new method for active learning, a subfield of machine learning or artificial intelligence. Active learning deals with the effective and efficient involvement of human annotators in a machine learning process. In addition to intelligently selecting data objects to train a classifier, MaPAL estimates the annotators' performances. Accordingly, annotators are preferably queried on data objects within their respective knowledge domains. MaPAL models the problem in a probabilistic and holistic approach to find the optimal solutions in a decision-theoretic sense. This approach showed a higher performance than comparable methods in the literature, as proven by statistical tests. The modeling approach is extensible to integrate further aspects of active learning in the future (e.g., different question types).