Explain Vector Embeddings to your mom

Ok ma! It’s simple i promise. sort of

the first step is to create tokens from large text - (assuming she already knows this)

right when we have large quantity of tokens, we convert them to numbers. why do we do so, simply because machines understand numbers better than text. now these numbers are kept in a structured format which can be termed as vectors or matrix which is used to represent vectors only.

What we essentially do is that we assign similar things similar numbers and different thing different numbers, now the computer knows that close numbers means the tokens that is represented by those numbers are closely related and vice-versa with different numbers different tokens are not related to each other.

Eg) Mom, Son —> 1, 1.1 - now since 1 and 1.1 are close we can establish that tokens represented by 1 and 1.1 are close.

this process is essential since machines cant understand emotions therefore we do this process that helps them relate to human emotions which we essentially call semantics - that is feeling/emotion/relationship between words that a machine understands