Definition
It is a hypothesis that states that words that are syntactically and semantically similar tend to appear near each other in a piece of text in the given language.
Using this hypothesis, word embedding (vectors containing information about these relationships) are learned by a Deep learning algorithm.