| Summary: | 碩士 === 國立臺灣大學 === 資訊工程學研究所 === 102 === Nowadays, personalization plays a crucial role in many software applications and in commercial advertisements. Different people have different views towards the world they perceive. By personalization, we can achieve a higher chance that people obtain the results they expect, which may further lead to improved user experience. Moreover, personalization also serves as an important basis for advertisements. Advertisement tailored to meet personal preference increases the likelihood that the target product or service attracts the attention of the target audiences, and therefore, may result in a higher product profit.
The goal of this thesis is to create a system that generates users’ Personal Ontologies based on their personal data, for example the history of their log queries and results selected after submitting these queries. These ontologies are used for refining users’ queries by keywords the users are interested in and that are semantically related to those queries. A graphical editor is also designed for viewing, editing and visualizing these ontologies. The visualization transforms the ontology into a graph. Multiple layout algorithms provided by Gephi [1] are supported. The editor can also be used for testing the query refinement.
This research is focused on three important aspects: ontology modeling, construction and evolution. Ontology modeling describes the model used in this approach and its notation. Ontology construction focuses on the way how the information is retrieved and saved in the ontology. Ontology evolution allows for changes in users’ preferences and reflects these changes in their ontologies.
The innovative approach to query refinement process is based on unique formulas, which use not only TF/IDF, but also employ time decay and word similarity metric so that the real meaning of the words contained in the ontology is taken into account. The thesis also shows strengths and abilities of the solution in several scenarios.
|
|---|
