NEREL

NEREL是一个大型的俄语数据集，专注于命名实体识别和关系抽取。该数据集由国立研究大学高等经济学院等多个机构合作创建，包含56000个命名实体和39000个关系，显著大于现有俄语数据集。NEREL的特点在于其对嵌套命名实体及其内部关系的标注，以及在语篇层面的关系标注。数据集内容丰富，包括事件、实体及其在事件中的角色等。NEREL的创建过程遵循最新的信息抽取方法和数据集标准，旨在解决知识图谱自动填充中的问题，特别是在处理嵌套和重叠实体的关系抽取方面。数据集的应用领域广泛，包括信息检索、自动文本摘要、问答系统和推荐系统等。

NEREL is a large-scale Russian-language dataset dedicated to named entity recognition and relation extraction. It was collaboratively created by multiple institutions including the National Research University Higher School of Economics and other partners. The dataset encompasses 56,000 named entities and 39,000 relational instances, which is notably larger than existing Russian-language datasets. A key characteristic of NEREL is its annotation of nested named entities and their internal relations, as well as discourse-level relation annotations. The dataset features rich content covering events, entities and their roles in events. Developed in accordance with cutting-edge information extraction methods and dataset standards, NEREL is designed to tackle challenges in automatic knowledge graph population, especially relation extraction involving nested and overlapping entities. It has wide-ranging application domains including information retrieval, automatic text summarization, question answering systems and recommendation systems.