FEEDBACK.
Minimalist black & white feedback platform with built-in NLP sentiment engine, TF-IDF keyword extraction, and radar chart metrics.
The Problem
Collecting feedback is easy, but extracting actionable signals (like primary complaints, satisfaction trends, or word associations) from text walls requires extensive manual effort. Heavy SaaS analytics tools are often bloated and expensive, creating a need for a lightweight, self-hosted option.
What I Built
I developed feedback., a clean, high-performance feedback management system. The application features a custom NLP text preprocessing pipeline using TF-IDF for keyword extraction and bigram association. It outputs a composite 0-100 index score based on sentiment weightings, plotted on an interactive radar chart.
Architecture Flow
A structural breakdown of the feedback ingestion and scoring logic:
Technical Decisions
I implemented the sentiment and TF-IDF extraction code directly in pure Python without importing heavy deep learning libraries like PyTorch or transformers. This minimized package sizes and kept API response speeds under 50ms, allowing it to compile statistics instantly.
Results and Impact
The minimalist dashboard presents feedback clearly:
- Sub-50ms Processing Speed for NLP extraction tasks.
- Intuitive Radar Plotting representing satisfaction vectors across five core business dimensions.