Kurdish Dataset for Fake News Detection (KDFND)

Famous Kurdish news websites, which is officially recognized by the Kurdistan Journal Syndicate and Facebook pages were used to scrape articles. Three separate Kurdistan cities are covered by the public websites. The Kurdistan Regional Government of Iraq's three cities are Erbil, Sulaimani, and Halabja. The news articles are written in Kurdish. This dataset is also notable for being the first and largest in the Kurdish language to concentrate on the Sorani dialect. Over the course of a year, articles were scraped daily from the preset public news sources using Python scripts. The articles were scraped using Facepager, Web Scraper, and Python tools. We eliminated all duplicate articles. Because there was no fact-checked platform in Kurdish at the time, articles from public news sources were also gathered from various news sources and social media pages. Each public news source was divided into two categories based on the annotation criteria for the articles in the dataset: Fake or Real. Each story that was labeled was given a designation based on its public source category, the Kurdistan Journalist Syndicate's guidelines for Kurdish journalism, and various additional criteria based on social media platforms.本数据集抓取了经库尔德记者联合会（Kurdistan Journal Syndicate）官方认证的知名库尔德新闻网站及Facebook页面所发布的新闻文章。该数据集覆盖伊拉克库尔德斯坦自治区的三座城市：埃尔比勒（Erbil）、苏莱曼尼亚（Sulaimani）及哈莱卜杰（Halabja），所有新闻文章均以库尔德语撰写。本数据集亦是全球首个且规模最大的聚焦于索拉尼（Sorani）方言的库尔德语新闻数据集。研究团队在为期一年的周期内，每日通过Python脚本、Facepager及Web Scraper工具从预设公开新闻源抓取文章，并剔除了全部重复内容。鉴于当时尚无经过事实核查的库尔德语平台，研究团队同时从各类新闻源及社交媒体页面采集了相关公开新闻文章。依据数据集的文章标注准则，每个公开新闻源被划分为“虚假”与“真实”两个类别；每篇已标注的文章均根据其所属公开源类别、库尔德记者联合会的库尔德新闻业规范，以及社交媒体平台的额外相关判定标准完成标注。

搜集汇总

背景与挑战

背景概述

该数据集是首个专注于库尔德语Sorani方言的假新闻检测数据集，包含从官方新闻网站和社交媒体收集的新闻文章，经过清洗和标记为真实或虚假两类，覆盖库尔德斯坦三个主要城市。