KazNERD
收藏资源简介:
KazNERD数据集是一个用于哈萨克语命名实体识别(NER)的数据集,包含从电视新闻文本中提取的112,702个句子,这些句子由两位母语为哈萨克语的专家在监督下进行手动标注。数据集采用IOB2标注方案,包含136,333个标注,涵盖25个实体类别。此外,数据集还提供了哈萨克语的标注指南和用于训练不同NER模型的代码(如CRF、BiLSTM-CNN-CRF、BERT和XLM-RoBERTa)。数据集的来源是电视新闻文本,大小为112,702个句子和136,333个标注,使用CC BY 4.0许可证,并提供了数据集的GitHub仓库链接。
The KazNERD dataset is a dedicated dataset for Kazakh named entity recognition (NER). It consists of 112,702 sentences extracted from television news texts, which were manually annotated under the supervision of two native Kazakh-speaking experts. The dataset employs the IOB2 annotation scheme, containing 136,333 annotations spanning 25 entity categories. Additionally, the dataset provides Kazakh-language annotation guidelines and code implementations for training various NER models, including CRF, BiLSTM-CNN-CRF, BERT, and XLM-RoBERTa. The source of the dataset is television news texts, with a total scale of 112,702 sentences and 136,333 annotations. It is released under the CC BY 4.0 license, and a GitHub repository link for the dataset is provided.




