Week2, prog_assgn_1, equalize

It is mentioned in the notebook:
β€œAs discussed earlier, if 𝑔g were defined by computing 𝑔1=π‘’π‘€π‘œπ‘šπ‘Žπ‘›βˆ’π‘’π‘šπ‘Žπ‘›g1=ewomanβˆ’eman; 𝑔2=π‘’π‘šπ‘œπ‘‘β„Žπ‘’π‘Ÿβˆ’π‘’π‘“π‘Žπ‘‘β„Žπ‘’π‘Ÿg2=emotherβˆ’efather; 𝑔3=π‘’π‘”π‘–π‘Ÿπ‘™βˆ’π‘’π‘π‘œπ‘¦g3=egirlβˆ’eboy; and so on and averaging over them, you would obtain a better estimate of the β€œgender””
My doubt is can we extract the gender feature vector from the embedding matrix E and use it as the estimate for gender?


I realize that its not correct to use the Gender feature from Embedding matrix as it only a proportion. Please ignore it.