community-datasets/urdu_fake_news
收藏资源简介:
该数据集名为Bend the Truth (Urdu Fake News),主要用于乌尔都语的假新闻检测任务。数据集包含两个主要字段:news(新闻内容,乌尔都语字符串)和label(标签,表示新闻是真实的还是虚假的)。此外,数据集还包含一个category字段,用于表示新闻的类别,包括体育、健康、技术、娱乐和商业五个类别。数据集分为训练集和测试集,训练集包含638个样本,测试集包含262个样本。数据集的创建和标注由专家完成。
This dataset, named Bend the Truth (Urdu Fake News), is primarily designed for Urdu fake news detection tasks. It includes two core fields: `news` (the news content, an Urdu string) and `label` (a tag indicating whether the news is real or fake). Additionally, the dataset features a `category` field that specifies the news category, encompassing five types: sports, health, technology, entertainment, and business. The dataset is split into training and test subsets, with 638 samples in the training set and 262 samples in the test set. The creation and annotation of this dataset were completed by experts.
数据集卡片:Bend the Truth (Urdu Fake News)
数据集描述
数据集摘要
- annotations_creators: expert-generated
- language_creators: expert-generated
- language: ur
- license: unknown
- multilinguality: monolingual
- size_categories: n<1K
- source_datasets: original
- task_categories: text-classification
- task_ids: fact-checking, intent-classification
- pretty_name: Bend the Truth (Urdu Fake News)
数据集结构
数据字段
- news: 一个乌尔都语字符串
- label: 指示提供的新闻是真实还是虚假的标签
- 0: Fake
- 1: Real
- category: 新闻的意图类别
- 0: bus (Business)
- 1: hlth (Health)
- 2: sp (Sports)
- 3: tch (Technology)
- 4: sbz (Entertainment)
数据分割
- train:
- num_bytes: 1762905
- num_examples: 638
- test:
- num_bytes: 799587
- num_examples: 262
数据集大小
- download_size: 1042653
- dataset_size: 2562492




