This repository hosts a filtered version of the great GermEval 2014 NER Dataset.
After some analysis of the annotated examples in this dataset, it can be seen that the dataset is highly biased by Wikipedia articles.
We present an overview of the top 10 top-level domains where annotations were retrieved from for training, development and test splits:
TLD
Number of examples (Percentage)… See the full description on the dataset page:
https://huggingface.co/datasets/stefan-it/germeval14_no_wikipedia.