News Impact Trainer — Emotion Labeling Tool

Web App + ML Pipeline

About

Interactive web application for training emotional intelligence models using news articles. Manual emotion labeling with keyboard shortcuts, persistent storage, and ONNX model export pipeline.

My Role

Built the full pipeline: RSS article aggregation, interactive labeling interface with 7 emotion categories, localStorage persistence, JSON export, and Python ONNX model export.

Core Capabilities

• Interactive emotion labeling with keyboard shortcuts (1-7 keys)
• RSS feed aggregation from multiple news sources
• 7 emotion categories: happy, angry, anxious, sad, neutral, hopeful, surprised
• Persistent localStorage with undo functionality
• JSON export of labeled dataset (763K+ articles)
• Python ONNX model export pipeline

Architecture

• Vanilla HTML5/JavaScript with no framework
• Node.js scripts for RSS feed fetching
• rss-parser library for article aggregation
• Python pipeline for ONNX model training and export
• localStorage for persistent label storage

Tech Stack