Our key contribution is to explain mathematically how word embeddings of Glove/W2V capture