Research intelligence for social media risk analysis
Detect Criminal Youth Recruitment Signals in Social Media Content
CrimeSentinel helps researchers, NGOs, and public-interest teams analyze Spanish-language social media content for recruitment, glorification, and safe civic communication signals.
"Join the family, here we have what you are looking for — respect, money, and brotherhood. Message me by DM if you are interested in the movement..."
In partnership with Civic A.I. Lab, Northeastern University & El Colegio de México
Sensemaking Pipeline
From Raw Content to Interpretable Risk Signals
CrimeSentinel organizes social media text through a structured workflow designed to support researcher judgment, contextual review, and careful interpretation.
- 1
Input Collection
Text, captions, comments, and contextual signals are prepared for analysis.
- 2
Context Pre-filter
Known civic, prevention, news, and government contexts are separated before classification.
- 3
Linguistic Signal Review
Spanish-language slang, recruitment cues, glorification terms, and coded phrases are evaluated.
- 4
Model Classification
The system classifies content into recruitment, glorification, or safe categories.
- 5
Human Interpretation
Researchers review outputs as decision-support signals, not final enforcement decisions.
Content Analyzer
Analyze Text for Recruitment and Glorification Signals
Paste a caption, comment, or short social media text to receive a structured classification output with confidence and contextual notes.
Input Text
Spanish-language captions, comments, or social media text
This tool supports human sensemaking. Results should be verified by a domain expert. The model achieves 95.8% accuracy on confirmed cartel content and requires human review for ambiguous cases.
Analysis Result
Classification, confidence, and contextual interpretation
Run an analysis to view classification, confidence, and contextual notes.
Research Capabilities
Intelligence Built for Social Media Safety Research
CrimeSentinel combines contextual filtering, linguistic signal review, and model-based classification to support careful analysis of online recruitment ecosystems.
Unified Signal Layer
Aggregate captions, comments, slang terms, civic context, and classification outputs into one structured research workflow.
Context-Aware Filtering
Separates civic, prevention, news, and government communication from risky content patterns.
Spanish-Language Signal Review
Supports analysis of coded phrases, slang, recruitment cues, and glorification language.
Human-Centered Interpretation
Frames outputs as decision-support signals for trained researchers, not automated enforcement.
Research Methodology
Built for Careful Analysis of Online Recruitment Ecosystems
CrimeSentinel is part of an academic research effort focused on understanding how criminal organizations use social platforms to normalize violence, recruit youth, and circulate coded language.
This tool is designed to support sensemaking, not replace human judgment. It helps analysts organize social media text into interpretable categories while preserving the need for cultural, linguistic, and situational review.
The system combines contextual pre-filtering, Spanish-language signal review, and classification outputs to help researchers identify patterns across recruitment, glorification, and safe civic communication.
Civic A.I. Lab | Northeastern UniversityEl Colegio de MéxicoResearch Orientation
Team
FAQ