YouTube Comment Sentiment Dashboard

NLP

TypeScript, React, VADER, YouTube Data API, Cloudflare Pages, Cloudflare Workers

A dashboard that scores a YouTube comment thread with VADER, a lexicon-based sentiment scorer, and shows the mix, the keyword spikes over time, and the comments behind each spike.

Problem

Reading a busy thread by hand is slow. The YouTube Data API hands comments over a page at a time under a daily quota, so the dashboard has to pull a whole thread inside that quota and score it fast enough to answer a browser request.

Context

lexicon, no training data
VADER