Hierarchical User Interest Modeling For Personalized Web News Service

With the rapid development of Internet technology, web data is increasing explosively. In the vast amounts of data, how to access the interesting information fast and efficiently gradually becomes the focus of users’attention. Web personalized service solves the contradiction between the rapid growth of web information and the relatively simple means to access information to a certain extent. Web information service providers attract users and improve user satisfaction by continuously improving the quality of personalized service.User interest modeling is the core of personalized services. It is applied in the fields of information retrieval, data mining, e-commerce and personalized recommendation to improve the quality of information services. Most of traditional user interest models are built on VSM (Vector Space Model) using keywords as the user interest. However, these models not only ignore the hierarchical granularity relations between keywords, but also ignore the use of domain knowledge to identify the specific concepts or topics of user interests. Thus, it is difficult to express the user interest accurately and reasonably in the user interest modeling.Motivated by the above problem, using personalized web news service as research background, we study hierarchical user interest modeling. We propose a Graph Partition-based Chinese Phrases Hierarchical Clustering algorithm called GCPHC. It organizes the user interest in a hierarchical tree, designs a HowNet-based Maximum Matching Mapping method called HNM3to map the user interest to the topics of ODP, and builds a hierarchical user interest model labeled with the topic for each cluster node. The main work of this dissertation is as follows:(1) In the context that web personalized service increasingly needs to be more intelligent, this dissertation uses a divisive hierarchical clustering algorithm to build a hierarchical user interest model for users interested web content, which is helpful for web personalized service;(2) For the problem of identifying the specific user interest topics in user interest model, this dissertation uses HowNet and ODP to map each hierarchical model node to the ODP topic, improving the effectiveness of the user model;(3) By the relevant experiments, we determine the suitable correlation function, data size and part of speech needed in the process of building the user model, which provides some basis for the application of hierarchical user interest in the field of web personalized service.