A One Hot Vector

one-hot-vectors

[0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0]

with one 0 for each word in a vocabulary where the 1 is representing the place of a word in the vector > In this kind of vector representation: none of the words are similar, they are all a 1.

Note that

Words are represented once in a vector. So words with multiple meanings, like "bank", are more difficult to represent. there is research to multivectors for one word, so that it does not end up in the middle (but you get already far by using 1 dense vector/word)

From: http://pad.constantvzw.org/public_pad/neural_networks_3