Module embeddings 

Post-training embedding access: similarity, analogy, save/load.

Usage

use word2vec::{Config, Trainer};

let corpus = vec!["the cat sat on the mat".repeat(50)];
let mut trainer = Trainer::new(Config { epochs: 3, ..Config::default() });
let emb = trainer.train(&corpus).unwrap();

let similar = emb.most_similar("cat", 3);
// [("mat", 0.92), ("sat", 0.87), ("on", 0.81)] (values illustrative)

// king - man + woman ≈ queen
// let queen = emb.analogy("king", "man", "woman", 3);

Structs

Embeddings

Trained embeddings with vocabulary — the primary inference interface.

Functions

cosine_similarity

Cosine similarity between two vectors (handles zero-norm gracefully).

normalize_vec

Return a L2-normalised copy of a vector.