Embedding Space Explorer

Open Embedding Space Explorer Fullscreen

This interactive MicroSim visualizes how text embeddings map phrases into a two-dimensional space where semantically similar phrases cluster together.

How to Use

1. Hover over any dot to see the phrase and its category.
2. Search by typing a phrase in the search box and pressing Enter or clicking Search. The simulator places your query in the space and draws dashed lines to the 3 nearest neighbors, showing Euclidean distances.
3. Reset clears the search and returns to the default view.

What This Demonstrates

• Semantic clustering: Phrases about AI/ML group together, cooking phrases group together, etc. — even though they share few words in common.
• Similarity search: When you enter a query, the nearest neighbors are the most semantically similar phrases — the same principle that powers RAG retrieval.
• Distance as meaning: Phrases that are close in the embedding space have similar meanings; distant phrases are unrelated.

Concepts Illustrated

• Embeddings
• Vector similarity
• Semantic search
• Nearest neighbor retrieval