| Property | Value |
|---|---|
| License | MIT |
| Author | QCRI |
| Downloads | 45,489 |
| Paper | Research Paper |
What is PropagandaTechniquesAnalysis-en-BERT?
PropagandaTechniquesAnalysis-en-BERT is a specialized BERT-based model designed for fine-grained propaganda detection in English news articles. Developed by QCRI, this model can identify specific propaganda techniques at both token and sequence levels, making it particularly valuable for media analysis and content verification.
Implementation Details
The model implements a multi-granularity neural network architecture based on BERT. It performs both token-level and sequence-level classification, utilizing the BertForTokenAndSequenceJointClassification architecture. The implementation supports the standard transformers library interface and can be easily integrated into existing NLP pipelines.
- Built on bert-base-cased architecture
- Supports both token and sequence classification simultaneously
- Implements 18 different propaganda technique classifications
- Available through the Hugging Face transformers library
Core Capabilities
- Fine-grained propaganda detection at fragment level
- Multi-class classification of propaganda techniques
- Joint token and sequence classification
- Real-time analysis of news articles
- Integration with existing NLP workflows
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its ability to perform fine-grained propaganda analysis at both token and sequence levels, offering more detailed insights than traditional document-level classifiers. It's specifically designed to identify 18 different propaganda techniques, making it highly specialized for media analysis.
Q: What are the recommended use cases?
The model is ideal for news media analysis, fact-checking organizations, content moderation systems, and research institutions studying propaganda and misinformation. It can be used to analyze news articles, social media content, and other text-based media for propaganda techniques.