%0 Journal Article %1 bakhtin2019learning %A Bakhtin, Anton %A Gross, Sam %A Ott, Myle %A Deng, Yuntian %A Ranzato, Marc'Aurelio %A Szlam, Arthur %D 2019 %K energy-models generative-models %T Real or Fake? Learning to Discriminate Machine from Human Generated Text %U http://arxiv.org/abs/1906.03351 %X Recent advances in generative modeling of text have demonstrated remarkable improvements in terms of fluency and coherency. In this work we investigate to which extent a machine can discriminate real from machine generated text. This is important in itself for automatic detection of computer generated stories, but can also serve as a tool for further improving text generation. We show that learning a dedicated scoring function to discriminate between real and fake text achieves higher precision than employing the likelihood of a generative model. The scoring functions generalize to other generators than those used for training as long as these generators have comparable model complexity and are trained on similar datasets.

@article{bakhtin2019learning, abstract = {Recent advances in generative modeling of text have demonstrated remarkable improvements in terms of fluency and coherency. In this work we investigate to which extent a machine can discriminate real from machine generated text. This is important in itself for automatic detection of computer generated stories, but can also serve as a tool for further improving text generation. We show that learning a dedicated scoring function to discriminate between real and fake text achieves higher precision than employing the likelihood of a generative model. The scoring functions generalize to other generators than those used for training as long as these generators have comparable model complexity and are trained on similar datasets.}, added-at = {2019-06-16T20:55:43.000+0200}, author = {Bakhtin, Anton and Gross, Sam and Ott, Myle and Deng, Yuntian and Ranzato, Marc'Aurelio and Szlam, Arthur}, biburl = {https://www.bibsonomy.org/bibtex/20ff8f68a6fab54094504f61b9882961b/patrickz}, description = {[1906.03351] Real or Fake? Learning to Discriminate Machine from Human Generated Text}, interhash = {10089afe0d5c53ee4879a550d8cabe73}, intrahash = {0ff8f68a6fab54094504f61b9882961b}, keywords = {energy-models generative-models}, note = {cite arxiv:1906.03351}, timestamp = {2019-06-16T20:55:43.000+0200}, title = {Real or Fake? Learning to Discriminate Machine from Human Generated Text}, url = {http://arxiv.org/abs/1906.03351}, year = 2019 }