Elicit user demands for item recommendation

Abstract

In an approach for eliciting user demands for item recommendation, one or more computer processors retrieve one or more items based on a user demand. The one or more computer processors update the one or more items based on the user demand. The one or more computer processors extract the one or more representative words corresponding to the one or more items. The one or more computer processors build a candidate item list based on the one or more representative words. The one or more computer processors generate one or more eliciting questions to help a user select an item based on the candidate item list.

What is claimed is: 
     
         1 . A method for user demand recommendation, the method comprising:
 retrieving, by one or more computer processors, one or more items based on a user demand, wherein the user demand is characterized by one or more demand vectors that are generated based on a model trained by reviews of the one or more items;   updating, by the one or more computer processors, the one or more items based on the user demand, further comprises:
 subtracting, by the one or more computer processors, the one or more demand vectors from one or more item vectors; and 
 and wherein the one or more items is characterized by the one or more item vectors and are generated based on a model trained by reviews of the one or more items; 
   extracting, by the one or more computer processors, one or more representative words corresponding to the one or more items by matching, the one or more item vectors with one or more representative word vectors and wherein the one or more representative words is characterized by one or more word vectors that are generated based on a model trained by reviews of the one or more items;   building, by the one or more computer processors, a candidate item list based on the one or more representative words, wherein the candidate item list is a KD-tree with the one or more representative words as one or more parent nodes and the one or more items as one or more leaf nodes;   generating, by the one or more computer processors, one or more eliciting questions to help a user select an item based on the candidate item list;   receiving, by the one or more computer processors, one or more user answers based on the one or more eliciting questions from the user;   determining, by the one or more computer processors, whether the one or more eliciting questions match the user demand based on the one or more user answers, further comprises:
 constructing, by the one or more computer processors, a user selection path, further comprises:
 determining, by the one or more computer processors, the one or more latest user demands and one or more corresponding user selection ranges; 
 generating, by the one or more computer processors, the one or more new user demands based on the previously determined user demand and the user answer wherein the user answer is based on the one or more additional eliciting questions; 
 generating, by the one or more computer processors, the one or more new user selection ranges; and 
 generating, by the one or more computer processors, the user selection path based on the one or more latest user demands, the one or more user selection ranges, the one or more new user demands, and the one or more new user selection ranges; 
 
 determining, by the one or more computer processors, whether the one or more range of a current node of the user selection path is above a predefined threshold; and 
 responsive to determining the one or more range of the current node of the user selection path is above the predefined threshold, determining, by the one or more computer processors, that the one or more additional eliciting questions does not match the user demand; 
   responsive to the one or more eliciting questions do not match the user demand, updating, by the one or more computer processors, the user demand based on the user selection path; and   responsive to the one or more eliciting questions matching the user demand, generating, by the one or more computer processors, one or more additional eliciting questions based on the candidate item list.