GloVe
A word embedding method that combines global co-occurrence statistics with local context windows.
Overview
GloVe (Global Vectors for Word Representation), developed at Stanford in 2014, is an unsupervised algorithm for learning word embeddings by factorizing the word co-occurrence matrix of a corpus. Unlike Word2Vec's prediction-based approach, GloVe explicitly leverages global statistical information.
Key Details
GloVe embeddings capture both syntactic and semantic regularities in language. The algorithm optimizes word vectors so that their dot product equals the logarithm of their co-occurrence probability. Pre-trained GloVe vectors (trained on Wikipedia, Common Crawl) are widely used as initialization for downstream NLP tasks.