Toward Extracting and Predicting Instance-Specific Attribute Values from E-Commerce Sites for Used Products

Abstract

When consumers purchase products through online, products information such as ratings, product reviews, product descriptions given by sellers are very useful for consumers to optimise their purchasing decisions. However, when a consumer purchases used products via online e-commerce sites, the consumer may consider much more attributes about the products than that for purchasing new products. This is due to the need for understanding instance-specific conditions before purchasing a used product and thus the available descriptions for a used product may differ in each other. In this paper, we proposed a design and implementation of a system that supports users to investigate instance-specific attribute values by extracting and predicting attributes and values of used items that are selling on e-commerce sites. Our key idea is preparing a system to identify instance-specific attributes as well as their values from the descriptions of items while browsing the e-commerce sites. Our system can also apply various machine learning methods to predict missing attributes values.