Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

VeriLens

VeriLens is an explainable, human-in-the-loop AI system designed to provide early misinformation risk signals for news articles and social media content.
Instead of determining absolute truth, VeriLens analyzes how information is written, how it may spread, and why it may require caution, supporting responsible human decision-making.


Problem Statement

Misinformation often spreads faster than verification.
Most existing solutions rely on post-hoc fact checking or binary true/false classification, which can be late, opaque, or overly authoritative.

There is a need for early warning systems that help users assess risk before content is widely shared.


Solution Overview

VeriLens provides a risk-based assessment of textual content using linguistic and behavioral signals.
The system emphasizes explainability, uncertainty awareness, and human oversight.

Key design principles:

  • Assistive, not authoritative
  • Explainable, not black-box
  • Risk-oriented, not binary classification

Core Features

  • Misinformation Risk Scoring
    Generates a calibrated risk score based on linguistic and contextual signals.

  • Explainability Signals
    Clearly explains why a piece of content may be risky (e.g., sensational language, missing attribution).

  • Virality Risk Estimation
    Assesses the likelihood of rapid spread based on language patterns.

  • Confidence & Uncertainty Modeling
    Displays confidence levels and uncertainty margins to avoid overclaiming.

  • Claim-Level Analysis
    Identifies the highest-risk claim within the content, when applicable.

  • Verification Signals (Optional)
    Provides non-authoritative cues that suggest when independent verification may be required.


Responsible AI Approach

VeriLens is designed in alignment with Responsible AI principles:

  • Does not determine factual truth
  • Avoids binary true/false labeling
  • Encourages human verification
  • Explicitly communicates uncertainty
  • Provides transparent reasoning

This system is intended as a decision-support tool, not a replacement for human judgment.


Technology Stack

  • Python
  • Scikit-learn
  • Natural Language Processing (NLP)
  • Explainable AI (XAI)
  • Streamlit

Project Status

This repository represents an MVP / prototype focused on: Core risk detection logic Explainability and transparency Responsible AI framing Future improvements may include: Expanded datasets Domain-specific calibration Optional integration with external verification services Bias and robustness evaluation Disclaimer VeriLens provides early misinformation risk signals based on linguistic and contextual patterns. It does not verify facts, determine truth, or replace professional judgment.

How to Run Locally

1. Clone the repository

git clone https://github.com/your-username/verilens.git
cd verilens
2. Install dependencies
pip install -r requirements.txt

3. Run the application
streamlit run app.py

About

Explainable AI system for early misinformation risk detection

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages