TaxoPublish: Towards a solution to automatically personalize taxonomies in e-catalogs

Taxonomies are utilized in e-catalogs to support customers navigating through a marketplace with the help of hierarchically structured concepts. However, when entering the e-catalog, each customer is shown the identical taxonomy regardless of individual requirements. Customers are distracted when navigating to preferred concepts as those are initially siblings of not required concepts. Provided progress in dynamic taxonomies, catalog segmentation, and personalized directories lacks in a fully automatic support for modifying the taxonomy according to the user’s actual requirements. The existing works need an explicit user-query, loose information about the domain, or require the modification through the provider. In this paper, TaxoPublish expert system based on logic programming is presented. The proposed system predicts the customers concept requirements for automatically modifying the taxonomy in B2B context. With TaxoPublish, retailers can now provide personalization in the form of personalized e-catalogs without any human effort, and without loosing any information about the domain. TaxoPublish is using knowledge provided through a CRM system for predicting customers preferences with the help of its semantic context weight, and knowledge of a PIM system for performing taxonomic operations based on two novel concept types. Through the usage of logic programming and the cross-plattform database model, TaxoPublish can be applied as expert system over distributed and heterogeneous data warehouse architectures across various domains to personalize and reduce the catalog without loosing semantics. The comprehensive experiments on two public and one private database show that TaxoPublish expert system is capable of fully-automatic taxonomy modification with an accuracy similar to the expert manual
modifications.