Export Reviews, Discussions, Author Feedback and Meta-Reviews

First provide a summary of the paper, and then address the following criteria: Quality, clarity, originality and significance. This paper addresses the classical problem of unsupervised learning of latent topic model, with an extra variable called response, which can be a score. The main issue is that the topic model (and the embeddings deduced from it) may not help in learning this extra variable, as the response can be induced by phenomena that are orthogonal to the topics. The goal of the so-called supervised topic model learning is to drive the topic learning into a direction which makes it useful w.r.t. the prediction of this extra variable (by a regression). The basic model considered is the Latent Dirichlet allocation (LDA) model.