GloVe (Global Vectors) is a method from Stanford for learning word embeddings from counts of how often words appear near each other across a whole body of text.
In plain words
GloVe first builds a big table of how often each pair of words occurs together across a whole body of text, then trains vectors so they explain those counts. The idea is that ratios of these co-occurrence counts carry meaning. The resulting vectors show neat patterns: the difference between man and woman is roughly the same as between king and queen. Stanford publishes the code and pre-trained vectors.
A quick example
Download a pre-trained GloVe file and load the vectors. Subtracting the "woman" vector from "queen" and adding "man" should land close to the vector for "king".
Where you'll meet it
- Pre-trained word vectors for classic NLP models.
- Teaching and studying word embeddings.
Related
Sources
- DocsGloVe: Global Vectors for Word Representation, Stanford NLP Group · read 29 Sept 2026
- PaperGloVe: Global Vectors for Word Representation, Pennington, Socher and Manning (ACL Anthology) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.