A Product Retrieval System Robust to Subjective Queries

Abstract: In this paper, we propose a method for retrieving products that match users ’ search queries written in natural language. We use a large amount of product reviews written by consumers, in order to match those queries with products. Since reviews include a lot of valuable information which commerce service providers don’t supply, by using these reviews our system can respond to a wide variety of users ’ requests, especially to subjective requests. If a review includes a description that coincides with the condition described in a user’s query, the product to which the review refers is considered the product that matches the query and is shown to the user by our system. Using our proposed method, we implemented an accommodation retrieval system named “Yado-tan”. The result of search experiments using natural language queries confirmed the retrieval capability of using our method