Signal-Based User Recommendation on Twitter Giuliano Arru, Davide Feltoni Gurini, Fabio Gasparetti, Alessandro Micarelli and Giuseppe Sansonetti Roma Tre University Via della Vasca Navale 79 Rome, 00146 Italy {arru,feltoni,gaspare,micarel,gsansone}@dia.uniroma3.it ABSTRACT In recent years, social networks have become one of the best ways to access information. The ease with which users con- nect to each other and the opportunity provided by Twitter and other social tools in order to follow person activities are increasing the use of such platforms for gathering informa- tion. The amount of available digital data is the core of the new challenges we now face. Social recommender sys- tems can suggest both relevant content and users with com- mon social interests. Our approach relies on a signal-based model, which explicitly includes a time dimension in the representation of the user interests. Specifically, this model takes advantage of a signal processing technique, namely, the wavelet transform, for defining an efficient pattern-based similarity function among users. Experimental comparisons with other approaches show the benefits of the proposed ap- proach. Categories and Subject Descriptors H.3.3 [Information Search and Retrieval]: [Information Filtering] General Terms Algorithms, Experimentation Keywords User recommendation, social network, twitter, signal pro- cessing, wavelet 1. INTRODUCTION As most of the current social platforms, Twitter 1 allows users to build networks of relationships. One user has the chance to start following someone, obtaining the new tweets of the followed person that will appear in the personal home- page of the website (i.e., timeline). However, the diversity and time-dependent evolving nature of user interests are not represented by the traditional social graphs of friendships and subscribers. In this paper, we propose a new approach of user recommendation on Twitter. It relies on a novel user model, called bag-of-signal, that allows us to represent how 1 https://twitter.com Copyright is held by the International World Wide Web Conference Committee (IW3C2). IW3C2 reserves the right to provide a hyperlink to the author’s site if the Material is used in electronic media. WWW 2013 Companion, May 13–17, 2013, Rio de Janeiro, Brazil. ACM 978-1-4503-2038-2/13/05. user interests change over time, so adding the time dimen- sion to user modeling in the Social Web. In order to fully exploit the potential of the new representation, we resort to mathematical tools used in the signal processing field, such as the discrete wavelet transform. 2. RELATED WORK Guy et al. [8, 3] propose a people recommendation engine within an enterprise social network site scenario. They ag- gregate several different sources to derive factors that might influence the similarity measure (e.g., co-authorships of pub- lications, patents or project-wikis). An extended analysis [3] proves the effectiveness of content-based approaches as op- posed to relationship-based algorithms, especially if histories of usage data in the social network are available. Further studies [9, 16, 15] show the benefits of tag-based profiles in the people recommendation task. Also Freyne et al. in [4] and Geyer et al. in [6] explore different recommendations strategies for improving the discovery of new users in so- cial networks and social media. Twittomender [10, 12] lets users find pertinent profiles on Twitter exploiting different strategies, both content-based and collaborative ones, once the user submitted an initial query of interest. Hannon et al. in [11] advance a faceted profile structure that makes different types of interest more explicit. None of the above approaches takes explicitly into account the time as rele- vant factor to include during the recommendation process. A first preliminary attempt of using the wavelet theory for the recommendation task has been proposed in [2, 5]. The authors suggest a comparison among time habits in order to improve traditional collaborative approaches for music rec- ommendation. 3. BAG-OF-SIGNAL MODEL The idea behind the approach we propose is to bring the problem of the user representation to the problem of the document representation, so allowing us to take advantage of the Information Retrieval (IR) techniques. Particularly, we drew inspiration from the work presented in [14]. In the context of the user recommendation on Twitter, some definitions are needed. We define pseudo-document related to a user u ∈ U the set of all the tweets t ∈ T posted by u in a given observation period: PD(u)= {t ∈ T | user(t)= u } where U is the set of all the users and T the set of all the tweets. A natural extension of the bag-of-word represen- tation is the bag-of-concept model, where concepts instead