Constructing Topic Hierarchy Tree for User Interests from Multiple Knowledge Bases

Many applications benefit from discovering and analyzing the topics of interest of social network users such as recommender systems. In this paper, we propose a methodology for building a topic hierarchy tree for user interests solicited from multiple knowledge bases. Different tree levels indicate different degree of abstraction, where topics are at the higher nodes and subtopics at the children nodes, i.e. leaf nodes are at the lowest level of abstraction. For each node, we aim to generate a diverse list of keywords that we call XWords lists; that contain list of words from which we can infer the node’s topic(s), we call these words: Topic Indicating Words (TIW). These TIWs are used for topic identification of users’ posts. To build the hierarchy we explore some of the available knowledge bases; namely, WordNet, Wikipedia and Directory Mozilla (DMoz) and integrate topics from those knowledge bases to build a complete topic tree for users’ interests.