reveal-nlp 0.3.0

Open-source Python library for intelligent harm detection and linguistic analysis in text

Reveal

An open-source Python library for intelligent harm detection and linguistic analysis in text.

Reveal combines Natural Language Processing, linguistic analysis, and cryptanalysis to identify harmful, coded, or anomalous language in text — helping researchers, platform moderators, and public safety professionals surface signals that standard tools miss.

What Reveal Detects

• Propaganda and persuasion techniques (22 categories)
• Grooming and predatory language
• Self-harm coded language
• Help signals from victims of abuse or bullying
• Geographic location indicators
• Emotional distress and sentiment
• Manipulative and accusatory tone
• Passive voice and agency obscuring
• Point of view shifts
• Shannon entropy anomalies
• Unusual n-gram sequences
• Statistically anomalous words

Installation

pip install reveal-nlp

Quick Start

from reveal.reasoning.reporter import analyze

text = "I feel so alone and trapped. Please someone help me."
result = analyze(text, fmt='text')
print(result)

Output formats

# Dictionary output for programmatic use
result = analyze(text)

# Formatted text report for human reading
result = analyze(text, fmt='text')

Project Structure

reveal/
├── harm/           # Harm detection, sentiment, reporting
├── linguistic/     # Voice, tone, POV analysis
├── cryptanalysis/  # Entropy, n-grams, anomaly detection
├── reasoning/      # Unified scoring and reporting
└── dictionaries/   # Word lists (JSON, community editable)

Contributing

Reveal is open source and actively seeking contributors — especially for expanding and validating the word lists in the dictionaries/ folder. No coding required to contribute word lists.

See docs/contributing.md for guidelines.

License

MIT License. See LICENSE for details.

Disclaimer

Reveal is a research and analysis tool. Word lists require ongoing community validation. Results should be interpreted by qualified professionals in appropriate contexts.