Ranking Amazon reviews by quality with pointwise ratings learned from pairwise data
Save your spotPowered by
No items found.
In this session, Tanmay shows how you can use TensorFlow to train a neural network to rank Amazon reviews by their "helpfulness" or "quality" automatically. Instead of training the networks on a regression problem, where we try to predict the helpfulness score directly, we will implicitly learn this score by learning to "choose" the more helpful review with an architecture inspired by IBM Research's Project Debater.
