Semantics

After training the neural network, to extract the word embeddings, you get the weight matrix $($the one between input layer and the hidden layer$).$ Assume that the vocabulary size is $10.$ Also, you used skipgram model to train the network.

Vocabulary

$\backslash$table$[[$man$,$machine,tower,capital,flight,grass,football,league,Germany,France$]]$

The weight matrix

$\backslash$table$[[0\mathrm{.}21,0\mathrm{.}47,0\mathrm{.}18],[0\mathrm{.}34,0\mathrm{.}55,0\mathrm{.}70],[0\mathrm{.}15,0\mathrm{.}17,0\mathrm{.}89],[0\mathrm{.}66,0\mathrm{.}1,0\mathrm{.}2],[0\mathrm{.}6,0\mathrm{.}38,0\mathrm{.}4],[0\mathrm{.}5,0\mathrm{.}27,0\mathrm{.}7],[0\mathrm{.}42,0\mathrm{.}17,0\mathrm{.}9],[0\mathrm{.}52,0\mathrm{.}56,0\mathrm{.}11],[0\mathrm{.}23,0\mathrm{.}51,0\mathrm{.}85],[0\mathrm{.}7,0\mathrm{.}4,0\mathrm{.}3]]$

According to the weights generated above, what will be the word vector corresponding to the word flight?

$(0\mathrm{.}6,0\mathrm{.}38,0\mathrm{.}4)$

$(0\mathrm{.}5,0\mathrm{.}27,0\mathrm{.}7)$

$(0\mathrm{.}66,0\mathrm{.}1,0\mathrm{.}2)$

$(0\mathrm{.}42,0\mathrm{.}17,0\mathrm{.}9)$