-
-
1st
-
2nd
-
3rd
-
4th
-
5th
-
6th
-
7th
-
next for this project
Inspiration
News anchors shape public opinion daily, yet their bias patterns often go unnoticed. We wanted to build a tool that makes media analysis accessible—letting anyone see the patterns in how stories are framed, what language is used, and which narratives dominate.
What it does
Bias Busters analyzes 90+ hours of news transcripts from Palki Sharma's "Vantage" show on Firstpost. It:
- Detects 5,150+ bias instances across 8 categories (Anti-China, Anti-Pakistan, Sensationalism, etc.)
- Timestamps every quote with direct YouTube links to verify context
- Visualizes patterns through interactive charts showing bias distribution
- Enables search across all transcripts to find specific topics or keywords
How we built it
- Data Collection: Scraped 92 YouTube video transcripts using Supadata API
- Backend: Flask API with keyword-based pattern detection and file caching
- Frontend: Vanilla JS with Chart.js 4.x for visualizations
- Analysis: Categorized content using propaganda keyword dictionaries
Challenges we faced
- YouTube transcript API rate limiting forced us to switch to Supadata
- Unicode encoding issues on Windows terminals
- Balancing between comprehensive keyword detection and false positives
What we learned
- The scale of repetitive narrative framing in news media
- How to build performant dashboards with vanilla JS (no React needed)
- Importance of timestamped evidence for credibility
What's next
- Add more anchors for comparative analysis
- Implement AI-powered sentiment analysis with Gemini
- Build browser extension for real-time bias alerts
Log in or sign up for Devpost to join the conversation.