Abstract

Functionality is of utmost importance to customers when they purchaseproducts. However, it is unclear to customers whether a product can reallysatisfy their needs on functions. Further, missing functions may beintentionally hidden by the manufacturers or the sellers. As a result, acustomer needs to spend a fair amount of time before purchasing or justpurchase the product on his/her own risk. In this paper, we first identify anovel QA corpus that is dense on product functionality information\footnote{The annotated corpus can be found at\url{https://www.cs.uic.edu/~hxu/}.}. We then design a neural network calledSemi-supervised Attention Network (SAN) to discover product functions fromquestions. This model leverages unlabeled data as contextual information toperform semi-supervised sequence labeling. We conduct experiments to show thatthe extracted function have both high coverage and accuracy, compared with awide spectrum of baselines.

 

Introduction (beta)

None

 

Conclusion (beta)

None