Meaningful Answer Generation of E-Commerce Question-Answering

In e-commerce portals, generating answers for product-related questions has become a crucial task. this article, we focus on the task of product-aware answer generation , which learns to generate an accurate and complete from large-scale unlabeled reviews product attributes. However, safe problems (i.e., neural models tend meaningless universal answers) pose significant challenges text tasks, question-answering is no exception. To more meaningful answers, in propose novel generative model, called Meaningful Product Answer Generator ( MPAG ), alleviates problem by taking reviews, attributes, prototype into consideration. attributes are used provide content, while can yield diverse pattern. end, generator with review reasoning module reader. Our key idea obtain correct question-aware information collection learn how write coherent existing answer. be specific, read-and-write memory consisting selective writing units conduct among these . We then employ reader comprehensive matching extract skeleton Finally, editor final question above parts as input. Conducted real-world dataset collected platform, extensive experimental results show that our model achieves state-of-the-art performance terms both automatic metrics human evaluations. Human evaluation also demonstrates consistently specific proper answers.