Optimizing small BERTs trained for German NER

Jochen Zöllner, Konrad Sperfeld, Christoph Wick +1

Currently, the most widespread neural network architecture for training language models is the so called BERT which led to improvements in various Natural Language Processing (NLP)…