Modeling Hierarchical User Interests Based on HowNet and Concept Mapping

Modeling user interests plays an important role in personalized service on the Internet. Many systems use classification methods to construct the user profile to represent user interests. However, it is difficult to cover all individual interests and capture the changes of user profiles. This paper introduces a method with two steps to solve the problem. In the first step it clusters the viewed web pages to extract individual interests by extending the vector of VSM with sememes from the HowNet. The second step is to map each interest cluster to the reference taxonomy to model hierarchical user interests based on similarity calculation of keywords and the extended vector. Experiments show that the method can model hierarchical user interests with a promising result. When a new interest emerges, it does not need any adjustment like collecting new training data or rebuilding the classifier. It can capture the diversified user interests and map to an interests taxonomy.