The Effect of Vectors on Semantics

A vector is like an arrow in a on a three dimensional graph that's one way you think of it right. In all these algorithms you you start with a randomly assigned collection of vectors so say there's 50,000 words. Each word gets a random vector and what happens over the course of this algorithm is that vectors start getting pushed towards each other based on which words are are near to each other. So at the end of training you now have words where the similarity of the associated vectors capture some kind of semantic meaning. And here's a very famous example. Let's look at this SAT style analogy that king is to queen what man is to woman.