Speaker
Description
In the talk, I will present the basics of Large Language Models (LLMs) and the status of the European LLM developments, in particular the OpenEuroLLM project that aims to develop open, multilingual foundational models in Europe and its challenges. I will then move on to summarise some recent developments on the other side of the research spectrum: symbolic meaning representations and latest developments of the most prominent ones, such as the Uniform Meaning Representation and the Prague Dependency Treebank. Such representations make heavy use of lexical resources – and I will show how these can be linked together and unified for making current and future meaning representation annotation efforts more consistent, while preserving all the information and alignment between form (plain text) and the representations. Towards the end of the talk, I will connect the current work on LLMs and meaning representations by postulating new research questions for which we now have resources and tools to work on.