Source-based-Fake-News-Classification

Context
Social media is a vast pool of content, and among all the content available for users to access, news is an element that is accessed most frequently. These news can be posted by politicians, news channels, newspaper websites, or even common civilians. These posts have to be checked for their authenticity, since spreading misinformation has been a real concern in todays times, and many firms are taking steps to make the common people aware of the consequences of spread misinformation. The measure of authenticity of the news posted online cannot be definitively measured, since the manual classification of news is tedious and time-consuming, and is also subject to bias.
Published paper: http://www.ijirset.com/upload/2020/june/115_4_Source.PDF
Content
Data preprocessing has been done on the dataset Getting Real about Fake News and skew has been eliminated.
Inspiration
In an era where fake WhatsApp forwards and Tweets are capable of influencing naive minds, tools and knowledge have to be put to practical use in not only mitigating the spread of misinformation but also to inform people about the type of news they consume.
Development of practical applications for users to gain insight from the articles they consume, fact-checking websites, built-in plugins and article parsers can
further be refined, made easier to access, and more importantly, should create more awareness.
Acknowledgements
Getting Real about Fake News seemed the most promising for preprocessing, feature extraction, and model classification.
The reason is due to the fact that all the other datasets lacked the sources from where the article/statement text was produced and published from. Citing the sources for article text is crucial to check the trustworthiness of the news and further helps in labelling the data as fake or untrustworthy.
Thanks to the datasets comprehensiveness in terms of citing the source information of the text along with author names, date of publication and labels.

背景
社交媒体是海量内容的汇聚池，在用户可访问的各类内容中，新闻属于最为高频的访问类型。新闻内容可由政客、新闻频道、报纸官方网站乃至普通民众发布。由于虚假信息传播已成为当下切实存在的公共隐患，诸多机构正采取举措提升普通民众对虚假信息传播后果的认知，因此需对上述发布的内容进行真实性核验。但在线发布新闻的真实性难以被精准判定，原因在于人工分类新闻不仅繁琐耗时，还易受主观偏见影响。

已发表论文
http://www.ijirset.com/upload/2020/june/115_4_Source.PDF

数据集说明
已针对《直面虚假信息》(Getting Real about Fake News)数据集完成数据预处理，并消除了数据偏斜问题。

创作初衷
在当下，虚假的WhatsApp转发内容与推特帖文足以影响缺乏辨识能力的受众，因此需将相关工具与知识落地应用：不仅要遏制虚假信息的传播，还要帮助民众认清其所浏览新闻的属性。面向用户的实用应用（旨在帮助用户理解其所浏览的文章）、事实核查网站、内置插件与文章解析器等工具可进一步优化，提升易用性，更重要的是，需强化公众对虚假信息的认知与防范意识。

致谢
《直面虚假信息》(Getting Real about Fake News)数据集在预处理、特征提取与模型分类任务中展现出了最优适配性。究其原因，其余所有数据集均缺失文章/陈述文本的生成与发布来源信息。标注文章文本的来源，对于核验新闻可信度至关重要，同时也能辅助完成虚假/不可信数据的标注工作。得益于该数据集具备全面的文本关联信息，涵盖文本来源、作者姓名、发布日期与标签类别。