How to convert predicted vector back to natural language?

[Tonyy123]

I have these vectors returned as part of running dmn_test.py after training. [array([14, 5, 8, 14, 14, 16, 20, 19, 16, 5, 8, 5, 8, 14, 16, 20, 16, 8, 8, 16, 8, 14, 16, 8, 5, 14, 16, 16, 5, 16, 20, 20, 14, 19, 19, 20, 16, 5, 14, 8, 14, 16, 19, 20, 20, 5, 19, 20, 8, 16, 20, 8, 19, 5, 19, 5, 19, 19, 16, 16, 19, 19, 14, 14, 19, 5, 16, 5, 14, 5, 14, 19, 16, 16, 8, 16, 8, 20, 8, 20, 16, 20, 8, 16, 16, 19, 5, 16, 20, 8, 5, 20, 16, 19, 16, 19, 19, 14, 8, 14])] [array([ 5, 19, 19, 19, 16, 19, 20, 16, 14, 14, 16, 5, 5, 14, 19, 20, 20, 19, 14, 20, 8, 8, 5, 20, 14, 20, 8, 20, 19, 20, 16, 5, 14, 20, 16, 5, 20, 5, 20, 16, 8, 8, 5, 14, 5, 20, 8, 16, 19, 20, 8, 16, 5, 20, 8, 20, 8, 20, 19, 19, 5, 8, 20, 14, 19, 5, 5, 19, 20, 14, 5, 5, 19, 8, 14, 16, 20, 5, 16, 8, 19, 14, 8, 8, 8, 8, 5, 16, 19, 20, 14, 20, 20, 20, 16, 16, 14, 14, 5, 8])] [array([ 5, 20, 5, 16, 16, 14, 19, 19, 5, 5, 20, 20, 19, 16, 5, 16, 5, 5, 20, 16, 14, 20, 20, 19, 8, 14, 20, 5, 16, 19, 19, 8, 5, 5, 16, 8, 8, 14, 20, 8, 8, 20, 8, 20, 19, 5, 5, 5, 19, 20, 16, 5, 5, 16, 20, 5, 5, 16, 8, 16, 14, 8, 19, 16, 8, 19, 14, 5, 20, 16, 19, 5, 5, 16, 16, 16, 19, 8, 5, 16, 5, 20, 19, 8, 19, 5, 16, 20, 19, 14, 19, 14, 14, 19, 20, 16, 5, 8, 19, 14])] [array([20, 19, 19, 5, 14, 16, 16, 5, 16, 8, 5, 19, 5, 19, 20, 16, 8, 8, 8, 8, 16, 5, 16, 16, 20, 20, 19, 8, 20, 14, 14, 20, 19, 20, 14, 19, 20, 14, 16, 8, 19, 14, 5, 14, 14, 14, 8, 5, 20, 19, 8, 16, 20, 16, 20, 20, 8, 16, 5, 20, 8, 16, 16, 16, 5, 19, 14, 8, 20, 20, 8, 19, 16, 16, 8, 20, 8, 16, 8, 20, 5, 16, 8, 5, 14, 20, 8, 5, 5, 8, 8, 16, 8, 16, 16, 20, 5, 19, 16, 5])] [array([ 5, 19, 16, 5, 16, 19, 14, 16, 19, 19, 20, 14, 8, 19, 14, 19, 8, 8, 14, 5, 19, 16, 16, 20, 16, 8, 5, 20, 19, 5, 19, 5, 8, 19, 5, 5, 5, 16, 14, 14, 14, 14, 5, 20, 20, 5, 19, 19, 8, 19, 8, 20, 19, 16, 19, 20, 5, 19, 16, 20, 16, 20, 8, 8, 14, 5, 19, 14, 19, 5, 19, 19, 14, 8, 19, 14, 19, 14, 20, 16, 5, 16, 8, 5, 16, 5, 16, 16, 19, 8, 16, 14, 20, 19, 20, 20, 20, 5, 8, 8])] [array([14, 8, 19, 5, 20, 8, 16, 19, 16, 5, 8, 5, 5, 5, 5, 8, 5, 5, 5, 20, 5, 20, 20, 8, 14, 14, 14, 14, 16, 14, 19, 19, 5, 19, 5, 14, 14, 14, 16, 16, 16, 19, 16, 14, 19, 19, 20, 8, 20, 19, 5, 14, 16, 20, 8, 16, 16, 19, 5, 5, 19, 5, 14, 8, 8, 14, 8, 16, 20, 16, 19, 19, 20, 14, 19, 19, 19, 5, 16, 14, 8, 8, 5, 20, 19, 14, 14, 8, 20, 14, 16, 5, 20, 14, 14, 8, 14, 8, 20, 19])] [array([ 8, 5, 19, 20, 16, 20, 16, 16, 20, 14, 5, 8, 14, 20, 20, 14, 14, 16, 8, 5, 20, 16, 16, 20, 19, 16, 19, 14, 20, 20, 8, 5, 16, 20, 20, 16, 8, 14, 20, 5, 8, 20, 8, 14, 19, 14, 14, 16, 14, 19, 14, 16, 20, 8, 14, 5, 20, 16, 16, 14, 5, 5, 8, 16, 14, 20, 14, 5, 8, 16, 8, 8, 20, 14, 19, 19, 16, 16, 8, 5, 8, 5, 19, 20, 20, 8, 19, 5, 5, 14, 19, 5, 16, 16, 16, 8, 20, 8, 20, 16])] [array([ 5, 19, 14, 14, 8, 20, 20, 5, 16, 14, 8, 16, 5, 5, 5, 8, 5, 19, 8, 14, 5, 14, 5, 16, 8, 20, 19, 20, 19, 14, 5, 8, 20, 16, 8, 16, 19, 19, 16, 20, 16, 5, 20, 14, 5, 16, 14, 16, 16, 5, 8, 14, 16, 20, 19, 5, 20, 19, 5, 14, 16, 14, 16, 20, 8, 16, 5, 14, 5, 19, 5, 16, 16, 14, 14, 16, 19, 16, 8, 19, 14, 14, 5, 19, 8, 5, 14, 16, 14, 5, 5, 5, 14, 20, 16, 20, 16, 8, 8, 8])] [array([19, 8, 14, 8, 16, 20, 8, 8, 16, 20, 16, 16, 14, 16, 8, 14, 16, 5, 16, 19, 14, 14, 19, 5, 5, 19, 14, 14, 8, 5, 5, 14, 14, 5, 5, 20, 20, 14, 5, 16, 8, 16, 8, 19, 14, 16, 16, 14, 16, 8, 16, 8, 5, 16, 16, 19, 8, 5, 19, 16, 14, 16, 8, 16, 8, 19, 8, 14, 20, 14, 14, 16, 16, 14, 16, 20, 16, 20, 16, 14, 5, 14, 20, 19, 20, 8, 5, 14, 16, 16, 5, 16, 16, 16, 8, 5, 8, 8, 5, 8])] [array([20, 16, 8, 16, 20, 16, 5, 20, 20, 14, 19, 19, 14, 16, 20, 20, 8, 5, 16, 16, 5, 14, 5, 14, 19, 14, 5, 8, 14, 5, 5, 16, 19, 8, 19, 19, 8, 14, 20, 16, 5, 19, 8, 8, 20, 20, 20, 20, 14, 19, 19, 20, 20, 8, 14, 20, 20, 5, 16, 16, 14, 8, 14, 8, 14, 19, 5, 5, 19, 5, 5, 5, 16, 8, 20, 8, 8, 14, 14, 19, 8, 19, 8, 5, 8, 5, 20, 14, 16, 8, 19, 20, 16, 14, 16, 8, 5, 5, 5, 19])] Could you please help to figure-out a way to convert these vectors back to natural language?

Thank you.

[Tonyy123]

For more clarification, those vectors are the values returned from this line from dmn_plus.py while running the dmn_test.py: pred = session.run([self.pred], feed_dict=feed)

[michaelxu1107]

hello, do you have any solution for convert predicted vector back to natural language?

[ainewsbot]

They are the index of the vocabulary i.e. ivocab. Simply get the value of that key. ivocab[pred[0]], which should give the answer in natural language.

[michaelxu1107]

They are the index of the vocabulary i.e. ivocab. Simply get the value of that key. ivocab[pred[0]], which should give the answer in natural language.

thanks

[michaelxu1107]

another question, i have trained model, but i find that the model only support answers which are one word, how to optimize model to support answers which are more than one words? thanks

[Maqsoodhuman]

They are the index of the vocabulary i.e. ivocab. Simply get the value of that key. ivocab[pred[0]], which should give the answer in natural language.

thanks

Did you got the value ?