Word2Vec Interactive Simulator
Skip-gram or CBOW with negative sampling. Watch the hidden-layer weights evolve from random initialization to final embeddings.
Corpus
the king is a strong man the queen is a wise woman the prince is a young man the princess is a young woman the man loves the woman the boy loves the girl paris is the capital of france rome is the capital of italy berlin is the capital of germany madrid is the capital of spain a dog is a loyal animal a cat is a small animal the dog chases the cat the king rules the kingdom the queen rules the kingdom
Hyper-parameters
Architecture
Skip-gram
CBOW
Embedding size (d)
Window size
Learning rate
Negative samples
Epochs
Speed (ms/epoch)
Seed
Initialize
Train
Step 1 epoch
Stop
Vocab size
–
Training pairs
–
Epoch
0
Loss
–
Inspect
Focus word (for neighbors & highlight)
Analogy: A − B + C ≈ ?
Compute
Hidden layer weights W (input → embedding)
Rows = words · Columns = embedding dims · Color = weight value (blue = negative, red = positive).
show numeric values
Hover a cell for its exact value.
2D projection of embeddings (PCA)
Words move as training progresses. Related words should cluster together.
Loss curve