scattertext

Why we included this project

Researchers and analysts who need to show why two text collections use language differently will get real mileage out of this one. It turns a corpus into an interactive HTML scatter plot, with each term plotted according to how strongly it leans toward one document group or the other, and labels are placed so they don't overlap. It handles Japanese text and works with parsing libraries like spaCy, so it slots into existing analysis workflows instead of living in a demo. The payoff is a self-contained HTML file you can open and explore in a browser, zooming and panning without setting up a server. For anyone comparing product reviews with support tickets, or speeches against press releases, it gives defensible visual evidence instead of a static word cloud.