AGRR-2019

AGRR-2019数据集是为俄语创建的，包含7.5k个带有gapping的句子和15k个相关负例句子。数据集涵盖了新闻、小说、社交媒体和技术文本等多种类型，旨在通过自动gapping解析共享任务推动NLP工具和方法的进步。创建过程中，使用了Compreno语法和语义解析器来预测句子中的null元素，并由11位评估者对标注进行评估，确保数据集的质量。该数据集主要应用于机器学习领域，特别是在处理语言中的省略现象方面。

The AGRR-2019 dataset was developed for the Russian language, consisting of 7.5k gapped sentences and 15k corresponding negative example sentences. It covers a wide range of text genres including news, fiction, social media content, and technical documents, and is designed to advance NLP tools and methods via shared tasks on automatic gapping parsing. During its construction, the Compreno syntactic and semantic parser was used to predict null elements in sentences, and 11 evaluators assessed the annotations to ensure the dataset's quality. This dataset is primarily applied in the field of machine learning, particularly for handling ellipsis phenomena in natural language.