The Knowledge Representation Learning (krAil) research group sees the Holy Grail of human-centered artificial intelligence in the machine learning of expressive knowledge representations.
Headed by Achim Rettinger the group develops technologies that exploit knowledge from human artifacts, specifically natural language in digital media and knowledge bases. Our goal is to build representation that are accessible to both, humans and computers while allowing expressive and explainable analysis and inference.
With our methods we analyze, interpret, and generate digital media that are dynamic, interactive and multimodal. We argue that digital media on a global scale is inherently multi-modal (visual, textual and conceptual) and multi-lingual and requires novel human perception models of media content.
To achieve that we research methods that combine:
- Machine Learning, specifically Deep Learning, Tensor-based methods and Probabilistic Models, with
- Symbolic Knowledge Representation, specifically Knowledge Graphs and Semantic Technologies
in order to enhance the inferencing capabilities of Machine Learning Algorithms and aid the construction of general-purpose, cross-domain symbolic knowledge bases.
We optimize, apply and evaluate our methods within specific application areas:
- Natural Language Understanding
- Computational Media Analytics
- Digital Humanities and Digital Libraries
- Knowledge Extraction, Integration and Refinement
