Skip-gram extensions: negative sampling, subsampling and phrase tokens

What changed

Extended skip-gram with negative sampling as a simplified noise-discrimination objective, frequent-word subsampling and data-driven phrase-token construction.

Scientific scope

Extended skip-gram with negative sampling as a simplified noise-discrimination objective, frequent-word subsampling and data-driven phrase-token construction. Main word2vec record explicitly covers the January2013four-author CBOW/skip-gram paper and excludes these extensions. This October2013five-author work adds Ilya Sutskever. Inherits distributional embeddings and NCE; NEG is not identical to normalized softmax likelihood, and empirical vector analogies are not general semantic reasoning guarantees.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Main word2vec record explicitly covers the January2013four-author CBOW/skip-gram paper and excludes these extensions. This October2013five-author work adds Ilya Sutskever. Inherits distributional embeddings and NCE; NEG is not identical to normalized softmax likelihood, and empirical vector analogies are not general semantic reasoning guarantees.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Distributed Representations of Words and Phrases and their Compositionality

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction