Responding E-commerce Product Questions via Exploiting QA Collections and Reviews

Proceedings of the 27th International Conference on Computational Linguistics, pages 2192–2203 Santa Fe, New Mexico, USA, August 20-26, 2018. 2192 1 Responding E-commerce Product Questions via Exploiting QA Collections and Reviews Qian Yu, Wai Lam, Zihao Wang Department of Systems Engineering and Engineering Management The Chinese University of Hong Kong {yuqian, wlam, zhwang}@se.cuhk.edu.hk Abstract Providing instant responses for product questions in E-commerce sites can significantly improve satisfaction of potential consumers. We propose a new framework for automatically respond- ing product questions newly posed by users via exploiting existing QA collections and review collections in a coordinated manner. Our framework can return a ranked list of snippets serv- ing as the automated response for a given question, where each snippet can be a sentence from reviews or an existing question-answer pair. One major subtask in our framework is question- based response review ranking. Learning for response review ranking is challenging since there is no labeled response review available. The collection of existing QA pairs are exploited as distant supervision for learning to rank responses. With proposed distant supervision paradigm, the learned response ranking model makes use of the knowledge in the QA pairs and the corre- sponding retrieved review lists. Extensive experiments on datasets collected from a real-world commercial E-commerce site demonstrate the effectiveness of our proposed framework. 1 Introduction Many E-commerce sites provide product information on product pages. Under a particular product page, users can ask questions about the corresponding product. Other experienced or qualified users can voluntarily provide answers. Thus, for each product, a product-specific question-answer (QA) collection is commonly available in E-commerce sites. However, the amount of QA pairs in the QA collection associated with a product is small since it is product-specific. In addition to the QA collection, users may write reviews about the product and a product-specific review collection is also commonly available. Figure 1 depicts an example, from a real-world E-commerce site, of a product page of a bluetooth car device. In this product page, the middle region contains the question-answer collection. The bottom region contains the review collection. Users may browse those two information sources for obtaining useful or insightful information about the corresponding product. Both information sources mentioned above would be helpful for providing instant responses to ques- tions newly posed by users about the corresponding product. Consider a new question “Is this device compatible with 2004 VW Jetta SportWagen?”, a semantically similar question can be found in the exist- ing QA collection of this product. Specifically the first question is a good match with the new question. Thus, the user posed answer for that question can be extracted as the response. On the other hand, con- sider another new question “Can I use iPhone 4s with it?”. This new question cannot match any existing question well. In spite of that, the review sentence “My iPhone 4S connected with the unit and uploaded my contacts with no issues.” found in the review collection of the corresponding product can provide informative content which can …

Content truncated to 500 words