What's in a name? Revitalizing the Named Entity Recognition Task

The classic named entity recognition (NER) task of identifying persons, organizations,locations, and more was immortalized almost 25 years ago in the CoNLL shared tasks.Much has changed since, but even more has stayed the same. This talk will explore threelines of research in NER: expanding NER from higher-resourced to less-resourcedlanguages, building NER models for e-commerce, and the difficulty of using LLMs for theNER task. I'll also discuss the common challenges encountered in all three lines ofresearch, such as evaluation reliability and the quality and availability of annotation.

Ponentes: Constantine Lignos