A common baseline that often works really well for sentence vectors is simply to average the word vectors for each word in the sentence. So lets say I have a vector made in this way. If I have a dictionary with lots of words and their associated word vectors, can I get the most likely sequence of words that made this sentence vector?

I'm fine with assuming we have 8 or less words in our sentence so unscrambling words is easy, but I'd like to have this algorithm work with millions of words in my dictionary.

  • 1
    $\begingroup$ You could train a recurrent neural network to do this. $\endgroup$ – Aaron Sep 22 '17 at 17:55
  • $\begingroup$ @Aaron what do you mean? $\endgroup$ – Phylliida Sep 22 '17 at 17:59
  • $\begingroup$ Ohh do you mean just use the decoder part of a seq to seq rnn? That’s a good idea $\endgroup$ – Phylliida Sep 22 '17 at 18:00
  • $\begingroup$ Yeah I think that's your best option. $\endgroup$ – Aaron Sep 22 '17 at 19:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.