Personalizing trending content in social media

Sha, Xiaolan
Thesis

Social media continuously draws the interest of researchers from a variety of perspectives - networks, sociology, marketing etc. In this networked age, the role of mass media at spreading information is increasingly opening itself to individual contributions. Researchers have therefore focused on how information is disseminated by individuals through social networks. Fluctuating along user connections, some content succeeds at capturing the attention of a large amount of users and suddenly becomes trending. Understanding trending content and its dynamics is crucial to the explanation of opinion spreading, and to the design of social marketing strategies. While previous research has mostly focused on trending content and on the network structure of individuals in social media, this work complements these studies by exploring in depth the human factors behind the generation of this content. We build upon this analysis to investigate new personalization tools helping individuals to discover interesting social media content. This work contributes to the literature on the following aspects:

o An in depth analysis on individuals who create trending content in social media, that uncovers their distinguishing characteristics;
o A novel means to identify trending content by relying on the ability of special individuals who create them;
o A mechanism to build a recommender system to personalize trending content;
o Techniques to improve the quality of recommendations beyond the core theme of accuracy.

Our studies underline the vital role of special users in the creation of trending content in social media. Thanks to such special users and their "wisdom", individuals may discover the trending content distilled to their tastes. Our work brings insights in two main research directions - trending content in social media and recommender systems.

 


HAL
Type:
Thesis
Date:
2013-05-06
Department:
Digital Security
Eurecom Ref:
3994
Copyright:
© TELECOM ParisTech. Personal use of this material is permitted. The definitive version of this paper was published in Thesis and is available at :
See also:

PERMALINK : https://www.eurecom.fr/publication/3994