◆ Data Science · Case Study 06

Search 80,000 UFO reports by meaning

Not keyword search. Type what you picture, and a neural network finds the real sightings that mean the same thing, even with different words. The model runs entirely in your browser.

← Back to the case study
🛸
How this works

A sentence-embedding model (MiniLM) is downloaded and run in your browser with WebAssembly, no server. It turns each real NUFORC report into a 384-number vector that captures its meaning, and does the same to your query, then ranks reports by cosine similarity. The index is built once and cached, so searches after that are instant.

Starting up…
Data: NUFORC via the TidyTuesday mirror · Model: all-MiniLM-L6-v2 via transformers.js (runs in-browser) · Built by Carlos Abel Vivanco / AbleV Labs
Back to the case study →