Free exploration
Select a word to inspect its semantic neighbours, similarity scores, and cluster relationships.
Embedding analytics
Nearest neighbours
Cosine similarity| Rank | Word | Category | Score | Distance |
|---|
Session progress
Learning guide
Semantic similarity
Words with similar meanings receive vectors pointing in comparable directions. Cosine similarity measures the angle between those vectors.
Dimensionality reduction
PCA preserves broad variation. t-SNE and UMAP-style layouts emphasise local neighbourhoods for easier visual exploration.
Vector analogies
Vector arithmetic can reveal relationships. A direction representing gender, place, or role may transfer between related words.
Context and ambiguity
A single vector struggles with words having several meanings. Contextual models create different representations for each sentence.
Bias awareness
Embeddings can reproduce patterns found in training data. Balanced examples and careful evaluation help reveal harmful associations.
Model differences
Word2Vec learns local prediction patterns. GloVe uses global counts. FastText adds subword information for unfamiliar or rare words.