Research On Recommender System Based On Semantic Web Technology

With the development of Internet, information resources on internet is increasing exponentially. It is difficult for people to find useful information in the world of extraordinarily large amount of internet information. There is inconsistency between the limitation, specificity, exclusion of information required by users and the dispersity and infiniteness of the internet information resources. In addition, the existing search engines can not provide good personalized service. Therefore, how to build user model by analyzing users' interests and preferences, and provide users with excellent personalized service is becoming the hot research topic.User model is the key technology in the personalized service system. Before the personalized services for different users with different interests and preferences are provided, user models for the personalized service system must be constructed. Most traditional user models use the keywords to describe the user's interests. However, the intrinsic relations among words is not considered in these methods. What's even worse, there is no the domain knowledge of the information resource that provides service for the construction of user interest model. In order to provide users with better personalized service, a recommender system based on semantic web technology is proposed in this paper, and the research emphasis is focused on the user model. The main research work in this paper is listed as follows.(1) A new type of user model is put forward by analyzing the user model. The model is constructed based on our interest classification ontology, each ontological user profile is obtained by instantiating the interest classification ontology. The spreading activation model in the field of the information retrieval is used for reference to complete the user model's renewal.(2) The traditional content-based recommender method is improved according to the presentation of the hierarchical user interests. The role of interesting value of every concept in the hierarchy of user interest is considered and the interesting value of each category concept is added in the users'interest vector by calculating, which can represent the users'interests better.(3) A recommender system based on semantic web technology, which is used help government departments to complete the work of public opinion analysis, is implemented based on the previous theoretical study. There are three modules in the framework of the system: news information collection and processing module,user model module and recommended strategy module. In the news information collection and processing module, the news and news comments is crawled from news sites to complete the information collection, and process the comments information of each news in a simple clustering method. The proposed ontology-based user model and updating methods is used in user model. Firstly,a interest classification ontology in the field of news is established, then the user profile is obtained by instantiating the interest classification ontology, and user model is updated by using the spreading activation model. The improved method of content-based is used in recommended strategy module. Finally, the proposed user model is verified based on the system.