BharatGather
A misinformation detection benchmark dataset focused on large-scale public gathering events in India. It contains 14,646 records, is constructed using a hybrid data-building pipeline, supports binary misinformation classification tasks, and incorporates cultural awareness as a design characteristic.
Timeline (last 90 days)
BharatGather is a multi-source dataset designed for binary misinformation classification (true/false determination) in large-scale gathering scenarios in India.
The BharatGather dataset contains a total of 14,646 records.
BharatGather was introduced in an arXiv paper (arXiv:2609.02895v1).
BharatGather adopts a hybrid data construction pipeline, comprehensively utilizing three methods: systematic web crawling, multimedia transcription extraction, and LLM-assisted synthetic augmentation.
BharatGather systematically scrapes verified true and false information from mainstream fact-checking platforms that has been validated by professional organizations.
BharatGather positions LLM synthetic augmentation as a means of ensuring narrative diversity, rather than as the primary source of data.
Multimedia transcription extraction is particularly important in the Indian context, as a large volume of rumors spread through non-textual formats such as short videos and voice messages.
BharatGather establishes a rigorous benchmark for evaluating the performance of detection systems in high-risk public environments.
All Facts (8)
BharatGather is a multi-source dataset designed for binary misinformation classification (true/false determination) in large-scale gathering scenarios in India.
50%UnverifiedThe BharatGather dataset contains a total of 14,646 records.
50%UnverifiedBharatGather was introduced in an arXiv paper (arXiv:2609.02895v1).
50%UnverifiedBharatGather adopts a hybrid data construction pipeline, comprehensively utilizing three methods: systematic web crawling, multimedia transcription extraction, and LLM-assisted synthetic augmentation.
50%UnverifiedBharatGather systematically scrapes verified true and false information from mainstream fact-checking platforms that has been validated by professional organizations.
50%UnverifiedBharatGather positions LLM synthetic augmentation as a means of ensuring narrative diversity, rather than as the primary source of data.
50%UnverifiedMultimedia transcription extraction is particularly important in the Indian context, as a large volume of rumors spread through non-textual formats such as short videos and voice messages.
50%UnverifiedBharatGather establishes a rigorous benchmark for evaluating the performance of detection systems in high-risk public environments.
50%