ukr-models/Ukr-Synth
收藏资源简介:
Ukr-Synth数据集是一个大规模的乌克兰语语料库,包含形态学标签、句法树和PER、LOC、ORG等命名实体识别标签。该数据集是从Leipzig Corpora Collection中抽取的乌克兰语报纸文本,经过句子分割和随机打乱后,使用基于Transformer的模型进行标注。数据集的语言为乌克兰语,格式为conllu,适用于命名实体识别、句法分析和词性标注等任务。数据集的大小在1M到10M之间,分为训练集和验证集,训练集包含1000000个样本,验证集包含10000个样本。数据集的创建基于Leipzig Corpora Collection,并遵循MIT许可证。
Ukr-Synth is a large-scale Ukrainian language corpus that contains morphological tags, syntactic trees, and named entity recognition tags such as PER, LOC, ORG. This dataset is extracted from Ukrainian newspaper texts in the Leipzig Corpora Collection. After undergoing sentence segmentation and random shuffling, it is annotated using a Transformer-based model. The dataset is in Ukrainian, with the CONLLU format, and is applicable to tasks including named entity recognition, syntactic parsing, and part-of-speech tagging. The overall size of the dataset ranges between 1M and 10M, and it is divided into a training set and a validation set: the training set includes 1,000,000 samples, while the validation set contains 10,000 samples. This dataset is constructed based on the Leipzig Corpora Collection and is released under the MIT License.
数据集概述
数据集描述
数据集总结
本数据集为一个大型的银标乌克兰语语料库,包含了形态学标签、句法树以及PER、LOC、ORG命名实体识别标签。该数据集是Leipzig Corpora Collection for Ukrainian Language的一个子样本,源文本为报纸文本,已被分割成句子并打乱顺序。句子通过基于黄金标准乌克兰语数据集训练的transformer模型进行标注。
语言
乌克兰语
数据集结构
数据分割
| 名称 | 训练集 | 验证集 |
|---|---|---|
| conll2003 | 1000000 | 10000 |
数据集创建
源数据
- 来源:Leipzig Corpora Collection
- 参考文献:D. Goldhahn, T. Eckart & U. Quasthoff: Building Large Monolingual Dictionaries at the Leipzig Corpora Collection: From 100 to 200 Languages. In: Proceedings of the 8th International Language Resources and Evaluation (LREC12), 2012
附加信息
许可信息
- 许可证:MIT License
- 版权年份:2022




