Hierarchical Attention Network (HAN) implemented in PyTorch
Aim is to predict review usefulness using the useful votes in Amazon reviews. Word and sentence attention weights are analysed to see if any insights can be drawn from the top attended words and sentences.
Code was written as part of a group project to predict and understand review usefulness through attention. Link to paper: https://drive.google.com/file/d/1yAbQExz45aMttI_waP6etlDzvF_EIbKm/view