PACIS 2020 Proceedings


Recently product reviews are used to answer questions on products, but the utility of social information such as review “helpful” votes for answering questions remains to be systematically investigated. In this paper, we model the problem of answering questions using product review “helpful” votes as a learning problem. We propose a helpfulness-aware mixture of experts (HaMoE) model for learning to rank relevant review sentences to answer questions. Based on the intuition that reviews with more “helpful” votes are more likely to contain useful information, we propose a parameterized helpfulness function of term features and learn to optimize their probability for answering questions. Experiment results on a widely used Amazon dataset show that our model significantly outperforms baseline approaches. Our main findings are: The helpfulness social information on reviews is helpful for answering questions. Specifically, modelling review helpfulness in terms of term features is effective to rank reviews for answering questions.



When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.