The same word, different meanings

Word2Vec has one important limitation: each stored word gets one fixed vector. But words shift meaning depending on where they sit.

bank by a river is not the bank you deposit money in. A bright light and a bright student are related but not the same. In a fixed word embedding, both senses share one starting point. That vector alone cannot say which meaning belongs here.

So the stored vector cannot be the final answer. It has to be a starting point that the sentence then adjusts. bank sets off from the same place every time, and the words around it, river or account, push it toward the sense in play. That adjusted version is a .

Which raises a harder question than it first appears. To let river adjust bank, the machine has to work out which of the surrounding words are worth listening to, and by how much. river matters here. the does not. And the word that settles the meaning can sit anywhere in the sentence, near or far.

Working that out is what the next module is about.