AI-Powered Social Crisis Alert Web System for Businesses

Inspiration: Why We Built This

One negative post can spiral into a full-blown crisis within hours...

In today’s digital era, social media acts as a megaphone for public opinion. A single bad experience can go viral and severely damage a brand’s reputation. That’s why we developed the AI Social Crisis Monitor to help businesses detect and respond to social media threats before they escalate.

What it does

Core Features:

Extracts Facebook Group (likes, comments, shares)
Sentiment analysis using AI (positive / negative / neutral)
Categorizes posts (e.g., reviews, issues, suggestions)
Summarizes key insights from each post
Includes source links to view original posts
Ranks top posts based on engagement and impact
Exports data in CSV for further analysis

Data Workflow: 1.Input brand name and Facebook Group link 2.Fetch latest group data via Supabase 3.Use Bright Data API to scrape posts 4.Auto-check status every 15 seconds 5.Pull post data in JSON (with post limit) 6.Filter posts mentioning the brand 7.Use OpenRouter AI to analyze sentiment 8.Extract insights, categories, and confidence level 9.Store all results in Supabase and Google Sheets

Output per post includes: Brand / Post link / Content Sentiment / Key insight / Category / Confidence

How we built it

This system was developed by integrating several key technologies to ensure efficient and accurate operation. These include:

AI (OpenRoute Model): For sentiment analysis, insight extraction, and comment categorization.
Bright Data API: For extracting posts from Facebook Groups.
Supabase and Google Sheets: For data storage and recording.
Automated Status Check System: To control the data extraction process.

Challenges we ran into

A significant challenge was data extraction from Facebook Groups, which faced limitations from the API and frequent changes in data structure, necessitating continuous adaptation and improvement of the data extraction methods.
We also encountered issues with Supabase connectivity, where some columns could not be fully extracted, requiring time to analyze and resolve database configuration problems.

Accomplishments that we're proud of

Built a 15-second interval status checker for data reliability
Integrated LLM via OpenRouter for smart and accurate sentiment + insight detection
Real-time connection to Google Sheets to enhance team collaboration ## What we learned
Flexible database design is critical for scale and maintenance
Proactive debugging and root-cause analysis are key to handling third-party API limitations

What's next for AI-Powered Social Crisis Alert Web System for Businesses

The future vision for this system is to evolve into an automated social media crisis detection and alert system, expanding its capabilities as follows:

Automated Crisis Detection: Advanced Sentiment Analysis: Hate Speech Detection Trending Topic Identification: Anomaly Detection: Identifies abnormal behavior, such as a sudden surge in negative comments.
Real-time Alerts: Sends immediate notifications through various channels like email, Line/Slack, Discord, and SMS (for severe crises), prioritizing alerts based on their severity level.
Comprehensive Dashboard: Displays a real-time overview of the situation (sentiment, hate speech), includes various trend graphs, offers in-depth categorization of comments, allows users to click for details of problematic comments, and provides data filtering and searching capabilities.
Early Warning System: Learns from past crises to identify common indicators, predicts future sentiment trends, identifies influential negative individuals or groups, and allows comparison of one's own situation with competitors.