Ban Competitors Scanner
The Ban Competitors Scanner uses advanced Named Entity Recognition (NER) to identify and manage competitor mentions in prompts. It helps maintain brand focus and prevents unintended competitor promotion in LLM-generated content.
Competitor detection workflow using NER model
Implementation Guide
Key Features
- Organization entity recognition
- Flexible competitor list management
- Optional name redaction
- Configurable detection threshold
- Context-aware scanning
Competitor List Best Practices
- Official company names
- Common abbreviations
- Brand variations
- Subsidiary companies
- Product line names
- Historical brand names
Configuration Options
- competitors: List of competitor names and variations
- redact: Boolean to enable name redaction
- threshold: Detection confidence threshold (default: 0.5)
- model: guishe/nuner-v1_orgs
Limitations & Considerations
- Model accuracy dependencies
- Context interpretation challenges
- Computational overhead
- List maintenance requirements
- False positive potential
Output Format
- sanitized_prompt: Text with optional competitor redactions
- is_valid: Boolean indicating if competitors were detected
- risk_score: Competitor detection confidence score (0-1)
Use Cases
- Marketing content generation
- Product documentation
- Customer support systems
- Brand protection tools
- Content moderation
Note: The effectiveness of the scanner heavily depends on the completeness of your competitor list. Regular updates and maintenance are essential for optimal performance.
Tip: Maintain your competitor list in a version-controlled configuration file. Consider implementing a regular review process to keep the list current with market changes.