In Cho's 2014 paper "Learning Phrase Representation using RNN encoder-decoder for SMT", they demonstrate the 2-D embedding of the learned phrase representation. I wonder how the phrase vectors can be quickly taken from a learned model? Is there any API or function to do this?
In Cho's 2014 paper "Learning Phrase Representation using RNN encoder-decoder for SMT", they demonstrate the 2-D embedding of the learned phrase representation. I wonder how the phrase vectors can be quickly taken from a learned model? Is there any API or function to do this?