Home▸Data Science▸Vector Space Search Engine: TF-IDF Document Retrieval

Vector Space Search Engine: TF-IDF Document Retrieval

Interactive vector space model: type a query and watch it get tokenized, weighted with TF-IDF, and ranked against a document corpus by cosine similarity in a live 3D projection of the vector space.

Data Science3DModerate60 FPS

Type a query and this simulator runs the exact pipeline behind classical text-mining and search-engine retrieval: tokenize the query, weight each term with TF-IDF (or plain term frequency), and rank a fixed corpus of ten documents by cosine similarity in a ~51-dimensional vocabulary space. A 3D scene projects every document and the live query into a spherical Fibonacci lattice so you can watch clusters of topically related documents pull the query point toward them, while the ranked list and inverted index on the left report the real, full-dimensional math driving the order.

⚙ Under the hood

Type a search query and watch it get tokenized, weighted with TF-IDF, and ranked against a document corpus by cosine similarity in a live 3D projection of the vector space.

text-miningtf-idfinformation-retrievalnlpvector-space-modelcosine-similarity

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install