.png){kind=small}
HinFakeNews
The HinFakeNews Dataset is a Hindi-language corpus of ~68,000 real and fake news articles across political, social, and communal domains.The HinFakeNews Dataset is a Hindi-language corpus of ~68,000 real and fake news articles across political, social, and communal domains. Real news is sourced from RNI-registered outlets; fake news from IFCN-approved fact-checkers. Each record includes URL, title, content, and a binary label (0=FAKE, 1=REAL).
About Dataset
The HinFakeNews Dataset is a comprehensive Hindi-language news corpus of articles spanning political, social, and communal domains. This dataset was created to support the development of approaches to effectively detect misinformation/fake-news spreading online. The news corpus consists of a collection of approximately 68,000 real and fake we-scraped articles, making it one of the largest available dataset for fake news detection tasks in Hindi. Real news articles are sourced from mainstream and Registrar of Newspapers for India (RNI) registered news publishing websites, while, fake-news are sourced from dedicated fact-checking platforms approved by International Fact-Checking Network (IFCN). The columns include the URL, Title of the article, Actual content and a Boolean value denoting whether the article was fake/real (0 - FAKE, 1 - REAL). The content of FAKE articles also includes a justification for the same alongside the actual piece of news. Hence, the use of this dataset is not just limited to fake news detection tasks, but also building explainable/evidence-grounded pipelines. Ethical considerations are strictly followed, with data sourced exclusively from publicly available fact-checking repositories, ensuring compliance with institutional guidelines on data privacy and responsible AI development.
Purpose of Dataset
Hinfakenews Is Designed To Address The Critical Gap In Low-resource Hindi Misinformation Detection By Providing A Large-scale, Credible, And Publicly Available Benchmark. The Dataset Supports The Development And Evaluation Of Fake News Detection Systems That Go Beyond English-centric Models, Enabling Research In Retrieval-augmented Verification, Linguistic Classification, And Explainable Ai For Indic Languages. Use Cases: Binary Text Classification, Where Supervised Models Such As Indicbert, Muril, And Xlm-r Are Trained To Distinguish Fake From Real Hindi News Articles. It Also Serves As A Corpus For Building And Benchmarking Retrieval-augmented Generation (Rag) Pipelines That Verify Claims Against A Hindi News Evidence Base. Researchers Can Use It To Evaluate Dense And Sparse Embedding Models On Hindi Text For Semantic Similarity And Retrieval Quality Tasks. The Dataset Further Supports Explainability Research Through Token-level Attribution Methods Such As Integrated Gradients, Stylometric Analysis Of Linguistic Cues Associated With Hindi Misinformation, And Cross-lingual Transfer Experiments From Multilingual Pretrained Models To Hindi-specific Classification.
Dataset Metadata
NA
India
Science, Technology and Research
Nitish S, Nitheesh, Rakshith Shenoy, Apoorva Kulkarni, Josh Boradia, Richa and Arti Arya
Structured
NA
NA
N.A.
14/06/26 10:53:26
Open
Hosted
Secondary
Web Scraping
Activity Overview
0
0 
95.83 MB 
48
Tags
- Text Analytics
- textual data
- Text Classification
- Fake News
- embeddings
License Control
NA
Select a file to preview its contents.
Data Quality Score Beta
Version Control
Version 1(95.83 MB)
- Arti Arya·4 day(s) ago
HinFakeNews-V1.xlsx
© 2026 - Copyright AIKosh. All rights reserved.