DEV Community

Cover image for Verisight: AI and Community-powered fake news detection
Mohamed Aashif
Mohamed Aashif

Posted on

Verisight: AI and Community-powered fake news detection

This is a submission for the Cloudflare AI Challenge.

What we have Built

Verisight uses AI and community efforts to help people assess the credibility of news articles. It evaluates the incongruency of articles by analyzing inconsistencies between headlines and the body, provides summaries, and allows users to add notes to provide additional context. It also cross-references articles through AI to show inconsistencies. Its primary objective is to empower users to avoid spreading misinformation.

Deployed Application - Verisight

Deployed Backend - Verisight Backend

Demo

My Code

Extension - Extension Repository

Backend - Backend Repository

Journey

As a team who browsed social media, we grew frustrated seeing news stories shared and later debunked. How could people know what to trust?

One evening, we discussed the rampant spread of misinformation online. No easy solutions seemed to exist. That's when the idea struck - what if we built a tool to verify stories in real-time?

We got to work brainstorming how it could work through AI and community-powered methods. We decided to create a browser extension that can summarize articles, detect inconsistencies, cross-check claims with other sources, and add notes with additional context.

Building the first prototype took many late nights of coding. But we were excited to launch a minimum viable product and get feedback. The journey was challenging, but rewarding to see users now have an easier way to evaluate news credibly in just a few clicks.

Our goal with VeriSight is to provide transparency and empower more informed decisions. There is still progress to make, and we look forward to continuing our work.

Task Types and Models Used:

Analysis and Incongruence:

  • Task: Text Classification
  • Model: @hf/thebloke/openhermes-2.5-mistral-7b-awq

Article Summarization:

  • Task: Text Summarization
  • Model: @hf/thebloke/openhermes-2.5-mistral-7b-awq

Cross-checking of Information:

  • Task: Retreival Augmented Text Generation
  • Model: @hf/thebloke/openhermes-2.5-mistral-7b-awq

Services used

  • Cloudflare Workers
  • Cloudflare Workers AI
  • Cloudflare D1

What We Learned:

  • Serverless Architecture: We gained valuable experience in implementing a serverless architecture using Cloudflare Workers, which provided scalability and cost-efficiency.

  • AI Model Integration: We learned the intricacies of integrating third-party AI models into a practical application, optimizing their performance, and ensuring seamless functionality.

My Team: @aashif @ppt1001 @ishmaifan

Looking forward to more challenges in the near future

Top comments (0)