Planeshifter/node-word2vec

Wraps Google's word2vec C implementation via child process execution, enabling efficient training of continuous bag-of-words and skip-gram models on text corpora. Provides semantic query methods including similarity scoring, nearest-neighbor search via cosine distance, and analogical reasoning (e.g., "king" - "man" + "woman"). Includes a preprocessing step (`word2phrase`) that identifies and concatenates frequently co-occurring word pairs, preserving multi-word concepts during vectorization.

354 stars. No commits in the last 6 months.