KI

The Institute's Data Science Lab is developing deep learning methods (artificial intelligence, AI) to automatically analyse sleep data at an expert level. This technology can not only improve our understanding of sleep and its role in various diseases, but also open up promising new diagnostic methods. It can also make it possible to conduct large, long-term studies that were previously not feasible. Our predictive models, which use EEG recordings to recognise sleep phases and features of sleep microarchitecture (e.g. sleep spindles) with high precision, are an important part of this development. With SomnoBot, we offer medical researchers the opportunity to use state-of-the-art neural networks to analyse sleep data, bridging the gap between research and application.

The Institute researches and develops large language models (LLMs) in various application contexts. Various embedding, prompting and training strategies are used to optimise the models.
The Data Science Lab 's research work includes the detection of toxic content in social networks (Facebook), the automated assessment of text complexity (Wikipedia), the automated analysis of speech acts in political debates (German Bundestag), and the development of technology to identify sexist statements in online comments (support for forum moderators).
Successes in international competitions, including twice as the best academic team at GermEval 2021 and 2022 and once as the second-best team at GermEval 2023, underline the role of our Institute in this rapidly growing field of research.