Project Hero
← Back to Portfolio

Links

Contents

📊 Tweefilter — Lightweight Tweet Filtering for Market Sentiment

The Problem: Financial Twitter Is Noisy

Financial Twitter is one of the fastest-moving sources of market sentiment, but it’s also full of noise. Traders looking for real-time signals often have to wade through low-value content, spam, or irrelevant posts—wasting time and missing key shifts in sentiment.

Existing tools are either too expensive, too broad, or not built with actionable filtering in mind.

The Solution: Simple, Targeted Filtering

Tweefilter is a lightweight tweet filtering system designed to capture and analyze market-relevant tweets using NLP. It parses incoming tweets, scores them for relevance and sentiment, and stores the most useful ones in a structured format—ready for use in trading workflows.

Built for cost-efficiency and speed, the system relies on IFTTT instead of the Twitter API, removing rate limits and credential management from the equation.

My Role: Solo Developer

I designed and built Tweefilter end to end, focusing on:

  • Cost-efficient tweet ingestion via IFTTT
  • Prompt-based NLP filtering using OpenAI
  • Sentiment tagging and relevancy scoring
  • Structured storage in Google Sheets for easy access

Key Features

📥 Tweet Ingestion via IFTTT

Instead of relying on the Twitter API, Tweefilter uses IFTTT webhooks to pull tweets in real-time based on keyword filters. This keeps the system free of authentication limits and avoids the cost overhead of premium API access.

🧠 NLP-Based Filtering

Incoming tweets are passed through the OpenAI API, which extracts sentiment, identifies tickers or market terms, and evaluates whether the tweet is useful for a trader. Tweets are scored and classified based on custom prompts designed around market context.

📄 Structured Output to Google Sheets

Only high-confidence tweets are stored in a connected Google Sheet. Each row includes:

  • Tweet text
  • Extracted tickers/keywords
  • Sentiment (positive, negative, neutral)
  • Relevancy score
  • Timestamp

This setup makes the results easy to scan, reference, or integrate into broader trading systems.

Tech Stack

  • IFTTT Webhooks: For tweet ingestion
  • Python