Heraclitus: A Framework for Semantic Web Adaptation

Heraclitus: Web Usage Driven Adaptation of the Semantic Web

Alexander MikroyannidisBabis Theodoulidis

School of InformaticsUniversity of Manchester

Introduction

The Semantic Web has emerged as a solution to the problem of organizing the immense information provided by the World Wide Web. However, a static Semantic Web can be of little use in the environment of the ever-transforming World Wide Web. The answer: Adaptation of the Semantic Web to the users’ needs and preferences.

Web Site Ontology (I)

It is strongly related to the site topology.It is comprised of the thematic categories covered by the site’s pages. These categories are the concepts of the ontology.The concepts are organized in a hierarchy, representing an “is a” relationship.The concepts are instantiated in the web pages.

Web Site Ontology (II)

Framework Principles

Web TransformationEnhancement of usability for all visitors, including

new onesTransparency

Tactical vs. Strategic adaptations (Coenen et al 2000)Emphasis on the role of the webmasterLearning adaptation engine

Adaptation of the physical and semantic structure: site ontology evolution

Architecture Overview

Topology & Ontology Evolution

Pagesets Classification

Session Mining

Preprocessing

PagesetsPagesets: : Sets of pages Sets of pages that are that are frequently frequently accessed accessed together together throughout throughout the same the same sessionsession

Preprocessing

Session identification approaches:TopologyContentTemporal information

Data Cleaning

Access Logs

Removal of:

Session Identification

Sessions

Accesses to multimedia

content Robot accesses

Erroneous accesses

Cleaned Access Logs

Session Mining

Market Basket AnalysisIncorporation of physical and semantic information: Web page

location Web page

classification

SessionsPagesets

GenerationPagesets

Web Site Topology

Web Site Ontology

Session Mining

Topology & Ontology Evolution

Pagesets

Linkage State

Classification

Content Classification

Web Site Topology

Web Site Ontology

Classified Pagesets

Refined Web Site Topology

Refined Web Site Ontology

Proposals Review

Report Generation

Report

Case Study

University of Manchester School of Informatics web site (www.informatics.manchester.ac.uk)2,500 web pagesApproximately 4,000 hits/day80% of the traffic is generated by undergraduate or postgraduate students

Web Site Topology Evolution (I)

Insertion of new shortcut links

Highlighting of popular existing links

Web Site Topology Evolution (II)

Web Site Ontology Evolution (I)New associations between conceptse.g.: Research and Programmes conceptsReorganization of concepts’ hierarchy. Creation of new categories, changes in others e.g.: Transfer of Staff concept to the highest level of the ontology New categorization of web pages. Identification of multiple instances of concepts or multiple subconceptse.g.: Job Vacancies page: categorized under Staff and Research

Web Site Ontology Evolution (II)

Conclusions

A web usage driven approach on the adaptation of the Semantic Web was introduced. The proposed framework targets both the physical and semantic aspects of the web.An architecture implementing the theoretical principles of the framework was proposed.Successful application of proposed methodology on a real web site.

Future Work

Automatic construction of the site ontology (e.g. agglomerative hierarchical clustering techniques) Meta-analysis of users’ access patternsSimultaneous adaptation of multiple web sites towards the development of the Adaptive Semantic Web

Thanks!

To try out Heraclitus visit:

http://heraclitus.sourceforge.net