Embedding Playground
Paste two sentences to compare them. Embeddings are computed in-browser with a deterministic feature-hashing model — instant and free, but lexical rather than fully semantic.
Sentences
Embedding settings
256
Higher dimensions give more signal for long text but add no cost here — this never leaves the browser.
Similarity
Cosine similarity
0.4221
Somewhat related
Cosine distance
0.5779
1 − cosine
Euclidean distance
1.0751
L2 on normalized vectors
Embedding dim
256
feature-hashed
Similarity gauge
Somewhat related
-101
Active features
Sentence A
43 / 256
Sentence B
54 / 256
Words and character n-grams are hashed into the vector. Sentences that share tokens land in overlapping cells, which is what produces a high cosine score.