Researchers have introduced TeViD, the first multimodal video fact-checking dataset specifically for the Turkish language. This dataset was compiled from articles by an IFCN-certified organization and augmented with translated content and synthetic examples to address class imbalance. Two verification pipelines were developed using open-source models: 3MFact-TR, which adapts a multi-role framework with Qwen models and includes custom modules for semantic checking and evidence retrieval, and a fine-tuned Qwen2.5-7B model that directly produces verdicts from claims and video descriptions. AI
IMPACT This work addresses a gap in low-resource language fact-checking, potentially improving the detection of misinformation in Turkish video content.
RANK_REASON The cluster describes a new academic dataset and associated research pipelines for multimodal video fact-checking in Turkish. [lever_c_demoted from research: ic=1 ai=1.0]
- 3MFact-TR
- Fine-Tuned Model
- International Fact-Checking Network
- LoRA
- Qwen2.5-7B
- Qwen2-VL-7B
- Snopes
- Turkish
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
