Add support for new ICDAR Europeana NER Dataset

[stefan-it]

Hi,

this PR adds support for our recently released ICDAR Europeana NER dataset.

The dataset itself is based on the French and Dutch part of the Europeana Newspapers NER dataset, and we performed further preprocessing steps, such as sentence splitting, punctuation normalization and introducing training/development and training splits.

The dataset is released with our "Data Centric Domain Adaptation for Historical Text with OCR Errors ICDAR 2021 paper and can be found in our repo here.

Usage

The dataset can be used in Flair like:

from flair.datasets import NER_ICDAR_EUROPEANA

french_corpus = NER_ICDAR_EUROPEANA(language="fr")
dutch_corpus  = NER_ICDAR_EUROPEANA(language="nl")

[alanakbik]

@stefan-it thanks for adding this!