In many artificial intelligence research areas, peak performance is achieved using large neural networks trained on vast amounts of unstructured data such as text or images. In addition to these data sources, many application areas additionally have manually structured, high-quality data in the form of knowledge graphs that have the potential to further enhance AI performance.
In this research area, we focus on integrating structured knowledge into neural networks, such as large-scale language models, and on using neural networks to automatically extract and improve structured knowledge from unstructured data.
In previous work, we have shown that language models trained on large amounts of unstructured data are able to estimate the reasonability of information contained in common sense knowledge graphs.
If you are interested in this topic, feel free to contact Janna!