deep learning - Is GAN a good (better ?) solution for text generation? -


since karpathy article rnn (http://karpathy.github.io/2015/05/21/rnn-effectiveness/) saw few projects , papers new evolutions in natural language processing found nothing text generation, nothing bring state generated text have meaning, on long term.

an saw gan projects on github, text generation. gans (better ?) way code text generator if we're looking general meaning long text ?

text genration open research topic , has been long time. in contrast gan research in infancy, many problems plaguing topic mode collapse, vanishing gradient , general difficult of training. not proven solution gan problem exists , if desireable.

still, gans have been applied text generation promising results (http://proceedings.mlr.press/v70/hu17e/hu17e.pdf), albeit not state-of-the-art results. question want do? want advance state-of-the-art in text generation? gans 1 direction can go advance field. if want generate text own application go more proven approach.


Comments

Popular posts from this blog

PHP and MySQL WP -

android - InAppBilling registering BroadcastReceiver in AndroidManifest -

go - golang pprof for c library code -