How did Google teach computers to understand that 'king minus man plus woman equals queen'?

The answer

In 2013, Tomas Mikolov and his team at Google published Word2Vec, a method for learning dense vector representations of words from massive text corpora using shallow neural networks. The model maps each word to a vector in 100–300 dimensional space, where semantic relationships emerge as arithmetic operations: the vector for 'king' minus 'man' plus 'woman' produces a vector closest to 'queen.' Trained on the Google News dataset of about 100 billion words, Word2Vec captured analogies, synonyms, and even cultural biases embedded in language. It became a foundational building block for virtually all subsequent NLP systems.