Learning to Generate Indistinguishable Product Reviews
Professor of Information Systems and the Leonard N. Stern Professor of Business
NYU Stern School of Business
Friday, May 4, 2018
10:30 AM – noon
Speakman Hall Suite 200
In this paper, we purpose a novel method called RevGAN to generate user reviews using a combination of Hierarchical AutoEncoder (hAE) and Conditional GAN (cGAN). We describe the proposed method and empirically demonstrate that it significantly outperforms several important benchmarks on the Amazon Review Dataset, and is also empirically indistinguishable from organic user reviews.