Word2Vec

Word2Vec is a group of neural methods that learns vector representations of words from the contexts in which they appear. The skip-gram approach predicts surrounding words from a target word, while continuous bag of words predicts a target from its context. Words used in similar settings receive nearby vectors, enabling similarity comparisons and simple semantic relationships. Word2Vec made embeddings practical for many language tasks, though modern contextual models represent the same word differently depending on its sentence. Its vectors can reproduce bias from the training corpus and handle rare, new, or ambiguous words poorly. Dataset choice, tokenization, dimensions, and evaluation therefore remain important.