NEREL: A Russian dataset with nested named entities and relations

NEREL is a Russian publicly available dataset for solving named entity recognition problem and relation extraction problem. The dataset contains more than 56K tagged entities and more than 39K relationships. An important difference between NEREL and previous datasets is the presence of markup for nested named entities.
The methods of extracting nested named entities differ from the methods of extracting "flat" named entities primarily by the architecture of the solution. Since NEREL provides annotations for nested entities, the article compared various approaches to solving this problem with the transfer to Russian language domain.