A solid paper with on an interesting and increasingly important problem: finding rationalizations/explanations for model predictions (in a sentiment classification task). The reviewers are all positive and after some discussion also fairly confident in their assessments.