The Hierarchical Understanding of Words

So basically without going into all the details you have these systems of vectors which interact with each other and they tell you what combinations to sort of take from them. So so like I was a query plus key equals value in some sense are those two first two variables can be transformed into value as kind of the answer. Or the query and the key assign a weight to the value. You're iterating this process what you do is you apply this attention over and over and over. And so what you're what you should be thinking of is that these words start looking at all the other words and start taking combinations of all the other Words until it builds up the hierarchical understanding.