word2vec-demo

w2v_demo.mp4

• This is a demo of word2vec skip-gram architecture
  • The sliding window size is 3
  • We use a center word (in one hot vector format) as an input to predict context words (in one hot vector format)
  • After the training is done, the hidden layer can be used as word embeddings and we can observe that similar words will be close to each other in vector space

How to run

• Open w2v.html by browser, and click Train button