CES Team Project Proposal

김기찬 | 김하윤 | 문다니엘 | 위예진 | 이신우 | 정은서

Target Customers

Our main target users are individuals interested in political and economic news who want to easily understand how leaders’ statements actually influence the market. Even without investment experience, anyone who wishes to grasp global trends can intuitively see “how words move markets” through our service.

Motivation

Currently, there are many platforms that summarize leaders’ speeches or analyze stock price movements, but few connect the content of speeches directly with market reactions. Most existing services either list news headlines or only visualize price charts without explaining the underlying cause.

We aim to bridge this gap by linking the language data of political and economic leaders directly with financial market data. Our platform allows users to intuitively experience how a specific speech affects the market. Unlike conventional analyst-focused platforms, we focus on a user-friendly interface and interpretation-centered visualization that make insights accessible to everyone, not just professionals. It is designed to be an open market intelligence platform that helps ordinary users understand the world through the power of language.

Primary Features

• Automatically collect speeches, posts, and interviews from major English-speaking leaders (U.S., U.K., Canada, Australia, etc.)
• Analyze sentiment (positive/negative), tone (hawkish/dovish), hedge level (uncertainty), and policy topics using NLP
• Automatically map each speech to relevant financial assets (indices, ETFs, currencies, commodities, etc.)
• Quantify market influence per speech and per leader through Power Rankings
• Provide visual dashboards for influence analysis, theme tracking, and impact decay curves

Additional Features

• Compare leader influence across countries, sectors, and policy topics
• Track market reactions to key themes such as inflation, AI regulation, and defense policy
• Provide real-time notifications for high-impact speeches
• Integrate social data (news mentions, tweets, media engagement) to enhance influence measurement

How to Run (Local Development)

1. Environment Setup

# Install Python dependencies
pip install -r server/requirements.txt

1.5 Run Database (Postgres + pgvector)

We use PostgreSQL + pgvector for team/production-friendly storage of entities, events, edges, and SBERT embeddings.

# From the project root
docker compose up -d db

Default DB settings (see server/db.py):

• DATABASE_URL: postgresql+psycopg://marketvoice:marketvoice@localhost:5432/marketvoice

If you want to override it, create server/.env:

DATABASE_URL=postgresql+psycopg://marketvoice:marketvoice@localhost:5432/marketvoice

2. Run Backend Server (FastAPI)

The backend server provides API endpoints for analyzed news data.

# Run from the project root
python3 -m uvicorn server.main:app --reload --host 0.0.0.0 --port 8000

• API URL: http://localhost:8000/api/news
• DB health: http://localhost:8000/api/health/db

3. Run Frontend Server (React)

The frontend is built with React (Vite).

# Move to the site directory
cd site

# Install Node dependencies (first time only)
npm install

# Run the development server
npm run dev

• Access the website at: http://localhost:5173