Basic Research in Deep Learning and Natural Language Processing

We investigate common tasks for classifying, extracting and linking textual data. Examples are entity recognition and linkage to Wikipedia (TASTY), linking text in a relational database (INDREX and IDEL), paragraph classification (SECTOR) or event/relation extraction.    

Explaining and Benchmarking

Understanding text-based information systems is another active research area. We focus on explaining common neural models, such as BERT. Moreover, we conducted research on benchmarking open information extraction systems (RelVis)

NLP & Healthcare: Understanding the Language of Medicine

Together with leading health care providers in Germany, such as Charité or Helios, and Health Care APPs, such as Ada Health, we investigate how we can leverage insights from text data, ontologies and tabular data for improving clinical decision making. Example are projects MACSS or Smart-MD.

Applying Natural Language Processing and Deep Learning

We apply our basic research on entity linkage or event extraction for supply chain management systems (projects Smart Data Web, PLASS and GoOLAP) or Market Analysis in the Fashion Domain  (H2020 FashionBrain). We also classify HatepSpeech (Project NOHATE) or design methods for detecting  bias in media.