On hierarchical web catalog integration with conceptual relationships in thesaurus

Abstract

Web catalog integration has become an integral aspect of current digital content management for Internet and e-commerce environments. The Web catalog integration problem concerns integration of documents in a source catalog into a destination catalog. Many investigations have focused on flattened (one-dimensional) catalogs, but few works address hierarchical Web catalog integration. This study presents a hierarchical catalog integration (EHCI) approach based on the conceptual thesauri extracted from the source catalog and the destination catalog to improve performance. Experiments involving real-world catalog integration are performed to measure the performance of the improved hierarchical catalog integration scheme. Experimental results demonstrate that the EHCI approach consistently improves the average accuracy performance of each hierarchical category.

Agent based Personalized e-Catalog Service System

“…Researches in recent years show that applying ontology to e-commerce scenarios would bring benefits such as solving the interoperability problems between different e-commerce systems [3,4]. Especially, eCatalog, which is a key component of e-commerce systems, seems to be the most adequate domain within e-commerce scenarios where ontology can realize the expression of eCatalog on a semantic level.…”

Improving Hierarchical Taxonomy Integration with Semantic Feature Expansion on Category-Specific Terms

“…In contrast, EHCI's hierarchical weighting mechanism considers the influences of category labels of more comprehensive neighboring levels on a document basis. The experimental results reported in Ho et al (2006) and Chen et al (2007) also show that EHCI is effective for handling large taxonomies. Therefore, we use EHCI as our baseline to study the effectiveness of the proposed SFE approach.…”

“…Previous integration research for hierarchal taxonomy integration mainly can be classified into two categories: clustering-based (Cheng & Wei, 2008) and classification-based (Ho, Chen, & Yang, 2006;Zhu, Yang, & Lam, 2004;Chen, Ho, & Yang, 2007). The clustering-based approach has the advantage in handling manifold taxonomies which may even have small overlaps and in performing integration without a priori training work.…”

“…Although the classification-based approach is more appropriate for handling taxonomies which have significant overlaps, it cannot handle the subtle relationships embedded in categories. For example, CatRelate uses five types of hierarchical relationships in a taxonomy to help catalog integration (Zhu, Yang, & Lam, 2004), and an integration scheme called EHCI uses a hierarchical weighting mechanism to strengthen the integration effectiveness (Ho, Chen, & Yang, 2006;Chen, Ho, & Yang, 2007). Nonetheless, CatRelate only discusses the hierarchical relationships on a category basis with a set of simple rules.…”

“…To study the effectiveness of SFE, we implemented it based on a hierarchical taxonomy integration approach (EHCI) proposed in Ho et al (2006) and Chen et al (2007) with the Maximum Entropy (ME) model classifiers. We have conducted experiments with real-world Web catalogs from Yahoo!…”