Word Embeddings

Count-based vs prediction-based — the full landscape with diagram

Sparse matrix problem, no semantic similarity, word order ignored

Two architectures explained intuitively with neural network diagrams

Vector arithmetic, 5-word vocabulary example, feature space visualization

Loading 300-dim vectors, most_similar, similarity, does_not_match methods

Count-based method that captures global co-occurrence statistics

Mathematical formula, geometric intuition, why not Euclidean distance

Comparing Word2Vec, Doc2Vec, and sentence embeddings