User Modeling on Twitter with Exploiting Explicit Relationships for Personalized Recommendations

Abstract

The use of social networks sites has led to a challenging overload of information that helped new social networking sites such as Twitter to become popular. It is believed that Twitter provides a rich environment for shared information that can help with recommender systems research. In this paper, we study Twitter user modeling by utilizing explicit relationships among users. This work aims to build personal profiles through a alternative methods using information gained from Twitter to provide more accurate recommendations. Our method exploits the explicit relationships of a Twitter user to extract information that is important in building the user's personal profile. The usefulness of this proposed method is validated by implementing a tweet recommendation service and by performing offline evaluation. We compare our proposed user profiles against other profiles such as a baseline using cosine similarity measures to check the effectiveness of the proposed method. The performance is measured on an adequate number of users.

Twitter User Modeling Based on Indirect Explicit Relationships for Personalized Recommendations

“…Also, the recent activities of most users do not have sufficient data or URL data to enrich their profiles from them. To address the problem of data shortage, the authors in [4] showed that direct explicit links between users should be used for the purpose of enriching the user profile. This finding will help expand the recent activities that are relevant.…”

“…They further stated that there is no agreement regarding who or what is an influential user. Subsequently, the authors in [4] argued that there is a need for creating a method which will be able to generate or produce an influence score based on the perspective of the normal user in relation to his or her interaction and behaviour. Their proposed influence score was higher when it identified influential friends of the user and then utilized those friends as a valuable source for building the user profile.…”

“…The usual way of building such a profile lacks up-to-date information about users' interests. The authors in [4] have solved this problem by using friends' recent tweets to enhance the user's profile with up-to-date information.…”

“…This paper is focused on modeling Twitter users through the exploitation of relationships with the aim of enhancing recommendations in the recommender systems. It is an extension of the previous work of the researchers of this paper [4] which was focused on direct explicit relationships (friends' accounts only), whereas this paper focuses on indirect explicit relationships. Indirect explicit relationships contain accounts that the user receives in their tweets from outside the friends list, such as friends of friends.…”

“…The methodology will contain a user profile designed or built from accounts that are outside the user's friends list. The proposed model makes use of the influence rule put forth in the authors' previous work [4], which will be explained generally in the following section. The influence rule will be used for the identification of influential friends.…”