January 2018
Beginner to intermediate
316 pages
7h 14m
English
In order to learn and extract word embeddings, Word2vec will implement another shallow neural network. This time, instead of generically throwing in new data into our visible layer, we will deliberately put in the correct data to give us the right word embeddings. Without going too into much detail, imagine a neural architecture with the following structure:

Like the RBM, we have a visible input layer and a hidden layer. In this case, our input layer has as many nodes as the length of vocabulary that we wish to learn from. This comes in handy if we have a corpus of millions of words but we wish to ...
Read now
Unlock full access