Meaning Map Lab
Semantic Comparison
Enter a few phrases, preview their embedding vectors, and watch them settle into a 2D meaning map. Similar phrases cluster. Different meanings drift apart.
One phrase per line. Use 3 to 8 phrases for the cleanest map. The vector preview shows only the first few and last few values.
Phrases
Edit the phrases directly. Every other panel reacts to these lines.
Similar ideas like “cat clinic” and “animal hospital” should move together. An unrelated phrase like “kitchen hose” should land away from that cluster.
Embedding previews
Each phrase becomes a high-dimensional vector. This preview shows a part of the list of embedding numbers.
2D meaning map
Nearby dots are semantically related. Labels are nudged apart to reduce overlap.
Drag inside the map to pan when points or labels extend beyond the visible frame.
The page embeds each phrase and then projects those embeddings into two dimensions for visualization.
Similarity heatmap
This matrix shows cosine similarity for every phrase pair. Darker cells indicate stronger semantic alignment.