Publications

Records are often approached positivistically, read as evidence of the information they contain; this applies both in an everyday context and in a scientific approach to records as data material. This qualitative interview study contributes to the epistemological reflection of records. Against the background of the analytical approach to records as artefacts, these are understood in terms of how they came into being. Using the example of records of appointment procedures for professorships at German universities, the study examines the question of which functions the records fulfil and how they are produced based on seminarrative expert interviews.

Online abuse in the form of online hostility and hate speech has become an increasing concern in digital publics, particularly affecting individuals engaging in political discourse. While research has extensively examined abuse on text-based platforms, little is known about how this phenomenon manifests on TikTok, a platform characterized by short-form videos, algorithm-driven virality, and participatory affordances. This study investigates the conditions and effects of online abuse on TikTok, focusing on German #PolitTikTok creators, who produce political content. Through, 21 semi-structured interviews, we explore (RQ1) how TikTok shapes online abuse, (RQ2) which factors make creators targets of online abuse, and ...

This report presents an overview of the benefits of implementing variable-level PIDs with regard to the FAIRness of data and offers a practical checklist outlining the requirements for service providers and the challenges faced by infrastructures aiming to adopt variable-level PIDs. In addition, it provides insights into the technical implementation of the PID test environment.

The financing of higher education in Germany involves a complex network of actors and institutions from the private and public sectors. Determining who contributes what to the financing requires considerable research and data processing as part of an in-depth structural analysis. Such an analysis for the year 2023 shows that the private and public sectors in Germany spent a total of around ¤59.806 billion on higher education (excluding research).

Nacaps, the National Academics Panel Study, is a new longitudinal study of doctoral candidates and doctorate holders in Germany funded by the Federal Ministry of Education and Research (BMBF). The project aims at providing nationwide cross-sectional and longitudinal data on doctoral candidates and doctorate holders in Germany regarding their study conditions as well as their career trajectories within and outside of academia. The Nacaps study series apply a panel design to multiple cohorts. Nacaps 2018 is the first cohort in this series of studies. In 2019, all doctoral candidates registered at 53 higher education institutions entitled to award PhDs/doctorates [...] Full Abstract: https://doi.org/10.21249/DZHW:nac2018:3.0.0

Nacaps, the National Academics Panel Study, is a new longitudinal study of doctoral candidates and doctorate holders in Germany funded by the Federal Ministry of Education and Research (BMBF). The project aims at providing nationwide cross-sectional and longitudinal data on doctoral candidates and doctorate holders in Germany regarding their study conditions as well as their career trajectories within and outside of academia. The Nacaps study series apply a panel design to multiple cohorts. Nacaps 2018 is the first cohort in this series of studies. In 2019, all doctoral candidates registered at 53 higher education institutions entitled to award PhDs/doctorates in Germany [...] Full Abstract: https://doi.org/10.21249/DZHW:nac2018:3.0.0

While Automated Machine Learning (AutoML) systems have shown strong performance on structured data, their application to natural language processing (NLP) tasks remains limited by static, task-agnostic search spaces. In this work, we propose a context-aware extension of AutoPyTorch that dynamically adapts both the hyperparameter search space and neural architecture configuration based on corpus-level meta-features. Our approach extracts interpretable textual statistics—such as average sequence length, vocabulary richness, and class imbalance—to guide the configuration of key hyperparameters. We also introduce two adaptive neural backbones, whose structures are shaped by these meta-features to improve model expressiveness and generalization.