Word Embedding Space Simulator

Word embeddings are the foundation every modern NLP system is built on: instead of treating words as arbitrary symbols, models like Word2Vec, GloVe and BERT map each word to a dense vector so that semantically related words land close together in the space. This simulator plots a small vocabulary as a live 3D point cloud — animals, royalty, cities and numbers each form their own visible cluster — and lets you pick any word to trace its nearest neighbors by cosine similarity. A built-in analogy button runs the classic "king − man + woman ≈ queen" vector arithmetic so you can watch the resulting point land right next to "queen," a direct visualization of why embeddings support reasoning about meaning, not just lookup.