Mobile Personalisation Enabled by Semantic Web Technology - Ensuring Quality in Ontology Building and Ontology Reuse

Abstract

Personalisation is the process that allows someone to receive services that fit the user’s particular needs. One of the main reasons to improve personalisation is the need for more control of personal information and the possibility for relevant services in an overloaded information society. There are many services available already that are not a success, and with a one size fits all approach, many services are not relevant for a user. While the goal would be to fully automate the personalisation process, we claim that we will come a long way with personal information available in personal profiles controlled by the user using open standards, Semantic Web technology and the possibility to combine different sources of information. This research is motivated by the belief that personalisation is an important feature for both users and service providers in a world of increasing number of individual mobile services. The main objective is to investigate use of Semantic Web technology for personalisation in mobile environments. Ontologies to represent personal information and domain knowledge are proposed in the personalisation architecture that will improve on the one size fits all approach that is often used. In this thesis we have considered personal profile information represented using Semantic Web technology and ontologies in particular. Before the ontologies have been created, we have developed frameworks to evaluate methodologies for ontology building. Such methodologies are useful to identify specific requirements and as a common grounding for people in a team that are to develop ontologies. Different groups can have differing requirements to what they need of support, and this is also considered in the frameworks. The developed evaluation frameworks have been used to evaluate a set of existing methodologies. The evaluation showed that the Ontology Development 101 methodology suited the requirements best. Ontology building methodologies mention reuse of existing ontologies as one step in the process. For our personal profile ontology we aimed to build on an existing ontology. However, there is no structural way to evaluate and decide whether an ontology satisfies the needs one has. Therefore we have created a framework that considers actual ontology requirements. This framework is used to evaluate candidate ontologies. The evaluation showed that reuse of existing ontologies is not a straight forward process. The methodologies for building ontologies have been useful in the ontology building process. Together with inspiration from existing ontologies, we have created a personal profile and food ontology. These are important sources of information that have been used in the proposed personalisation architecture. For the development of the personal profile we have used developed persona and scenarios as a basis. An important part of the personalisation architecture is the mediator. A prototype that covers parts of the mediator that uses profile and domain information has been implemented. The prototype is implemented in Java, using OWL API, Jena API and Pellet API. The implementation based on test-driven development has been evaluated using the persona and scenarios and shows that the results from the personalisation are comparable with the expected results. To evaluate the personalisation method we have used the Technology Acceptance Model based on a questionnaire and two videos created from the persona and scenarios, one for a push service and one for a pull service. The survey found that many of the participants were positive towards the proposed personalisation. Perceived usefulness was the strongest predictor for use for both the push and the pull service. We achieved a promising result with the use of ontologies and Semantic Web technology for a personal profile. The personal information in a profile ontology is useful in the personalisation process. One can programmatically make use of the profile and related domain information in the personalisation process to retrieve a personalised result.