Deep attention-based model for helpfulness prediction of healthcare online reviews

With tons of healthcare reviews being collected online, finding helpful opinions among this collective intelligence is becoming harder. Existing literature in this domain usually tackled helpfulness prediction with machine-learning models optimized for binary classification. While they can filter out a subset of reviews, users might be still overwhelmed if the number of reviews marked as helpful is high. In this paper, we design a new neural model optimized for predicting a continuous score that can be used to rank reviews based on their helpfulness. Given embedding representations of words in a review, the proposed model processes them through recurrent and attention-based layers to solve a helpfulness prediction task, modeled as a regression. Experiments on a real-world healthcare dataset show that the proposed model optimized for regression leads to accurate helpfulness prediction and better helpfulness-based rankings than models optimized for binary classification.