The Dynamics of Content Popularity in Social Media

Introduction

The startling success of Web 2.0 applications during the last years has reshaped our views on how information is generated and distributed to the masses. While traditionally the content items that appeared in the web pages of online publishers (e.g. news sites) were selected by a single authority or a small committee of experts, the advent of Web 2.0 applications, such as the Social Bookmarking Systems (SBS), has allowed mass participation in the content selection process.

In an SBS, users upload and save links to web pages (bookmarks) that they deem interesting. These bookmarks are usually public: Once they are submitted by a user, the rest of the community is able to view them and if they consider them interesting, they may give a 'thumbs up', i.e. rate them. Thus, applications of this kind, originally meant to provide a convenient tool to organize one's bookmarks, have evolved to act as a content popularity (or 'interesting-ness') ranking mechanism. Examples of such systems are digg¹, propeller² and newsvine³. The proliferation and success of such applications has created online communities where the bookmarked content items undergo a perpetual rating process by the community. Typically, the most popular content items coming out of this process are displayed in a prominent place at the application’s site. In that way, the vision of social media is realized.

The work presented in this paper has been mainly motivated by the fact that the content rating processes, such as the ones taking place within social media applications, have not been sufficiently studied so far from the temporal and social point of view. For this reason, we established a framework for the study of such social media aspects and applied it for analyzing the content rating dynamics emerging in the digg application. More specifically, through the work described in this paper, we make the following contributions:

• •

  Introduce the Diggsonomy framework, inspired by the Folksonomy framework of (Mika, 2005) and (Hotho et al., 2006a), to enable the study of online content popularity;

• •

  Collect and analyze the statistical properties of an extensive data set from digg;

• •

  Provide insights to the temporal evolution of popularity and to the role of the social relations on the content rating process.

We consider the work presented here to be of significance for the following applications:

• •

  Administration of SBS and social media applications. Since the operation of such a service involves a large investment in server and network equipment, better understanding of page popularity and thus request patterns could contribute to smoother service provision and/or reduced infrastructure costs by means of efficient data partitioning and page caching schemes.

• •

  Recommender system development. Understanding of the temporal and social dynamics of content popularity can be exploited to improve the precision and relevance of recommendations for systems where these aspects are of particular significance (e.g. news articles, advertisements).

The rest of this paper is structured as follows. Section 2 provides a short overview of existing work in the fields of mining web 2.0 data and analysis of related temporal and social phenomena. In Section 3 the proposed social media analysis and modeling framework is introduced. Further, the data collection and the associated experimentation as well as a discussion of the findings are provided in Section 4. Finally, Section 5 concludes the paper.