nlp-toolkit
por raghulpasupathi
Search & ResearchResumen
Advanced NLP with perplexity scoring, burstiness analysis, and entropy calculation.
Descripción
Qué hace esta skill
Analyzes text using statistical NLP measures to estimate whether it was written by a human or generated by AI. Computes perplexity, burstiness, and Shannon entropy scores, then combines them into a confidence rating.
Runs entirely client-side with no external API calls, so text is never sent to a third-party detection service.
Cuándo usarla
- Screening submitted essays for AI authorship
- Flagging AI-generated support tickets before human review
- Auditing content pipelines for model-generated text
- Comparing writing samples before and after AI tool adoption
- Spot-checking blog drafts for undisclosed AI use
Flujo de trabajo de ejemplo
User asks: Is this 200-word paragraph AI-generated?
- 1
Pass the text to analyzeText with default thresholds
- 2
Calculate perplexity from unique-to-total word ratio
- 3
Measure burstiness as coefficient of variation across sentence lengths
- 4
Compute Shannon entropy over character frequency distribution
- 5
Combine three scores into a single confidence percentage
{ isAI: true, confidence: 78, metrics: { perplexity: 32.4, burstiness: 0.21, entropy: 4.1 }, explanation: 'Low perplexity and uniform burstiness suggest AI generation' }
Instalación
clawhub install raghulpasupathi/nlp-toolkitEjecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.