User Personalized Services Based on Semantic Web for Pervasive Computing

In pervasive computing environment, more personalized information can be achieved, and better user profile model can be built, thus personalized service can be realized. User profile is the critical aspect in personalized service of digital library. This paper proposes structure and mechanism of user profile in personalized service of digital library. The main problems during its construction are hereinafter introduced in details, which can bring pervasive computing and Web service together tightly. Some technical problems about pervasive-computing-based personalized service are discussed as well include representation, creation and update of user profile. User profile is represented by vector space model (VSM) based ontology. Simple domain ontology of digital library is built first, and then user´s interests computed by VSM, in which the item of vector is ontology or concept about user´s interests. Features are selected by combining classifying and clustering algorithm, and take support vector machine (SVM) and unsupervised clustering (UC) as example to show the implementation. Using hybrid model, which contains long term model and short term model, compute the interest drifting of user. These two models use different forgetting strategy: short term model use time window and long term model use gradual forgetting algorithm.