RELATIONSHIPS IN SOCIAL NETWORKS REVEALED:
A FACEBOOK APP FOR SOCIAL GRAPHS IN 3D
BASED ON X3DOM AND WEBGL
Nikita Mattar and Thies Pfeiffer
A.I. Group, Faculty of Technology
Bielefeld University, Germany
ABSTRACT
From the perspective of individual urs, social networking platforms (SNPs) are meant to reflect their social relationships. SNPs should provide feedback allowing urs to exploit the information they have entered. In reality, however, most SNPs actually hide the rich social network constructed by the urs in their databas behind simple ur interfaces. The interfaces reduce the complexity of a ur's social network to a text-bad list in HTML. This article prents results from a ur study showing that 3D visualizations of social graphs can be utilized more effectively – and moreover – are preferred by us
ers compared to traditional text-bad interfaces. Subquently, the article address the problem of deployment of rich interfaces. A social graph application for Facebook is prented, demonstrating how WebGL and HTML5/X3D can be ud to implement rich social applications bad on upcoming web standards. KEYWORDS
3D Graph Visualization, Interaction, JavaScript, Social Networks, WebGL, X3D/X3DOM.世界名牌化妆品排名
1.INTRODUCTION
Online social networking platforms (SNPs) are, as of 2010, very popular. Millions of urs have entered their profiles and, at least, defined their relationships, if nothing el. This nsitive information is collected in large databas and it is in principle easy to digest by data mining. There exists a cornucopia of scientific tools to analyze and visualize social networking data (e Related Work) on a large scale. However, besides the ongoing highly relevant discussion about the potential misu of this data by third parties, e.g. regarding privacy issues in the large, one aspect has, to our knowledge, been neglected by most platforms: individual urs are not empowered to get a comprehensive overview of their social network in the small scale. Thus, the data entered by the urs is only partially reflected back to them. Most social networking platforms simply lack prentati
ons beyond rialized lists of first grade relationships. While this approach is suitable for finding special known entities in one’s social network, it does not provide a straight answer to questions like, e.g., which of one's friends are also friends or what common interests they share, and it makes it difficult to discover something new in the structure of the network.
From the different alternatives regarding the lection of features to prent to urs and the way they are prented, this paper argues to u graph-bad reprentations for the basic relations expresd in the network. Rearch on social network analysis and graph-bad systems in general has developed a rich t of tools and visualizations which could be ud to fill this gap. Visualizing social networks as a graph emphasizes the structure of the data and allows for an intuitive exploration.
Graph-bad visualizations for social networks can be dated back to the early 20th century. But why, to the knowledge of the authors, does not any of the current social networking platforms u graph-bad visualizations? In this article, we investigate three possible reasons: (a) the tasks a typical ur performs on such a platform do not require the visualizations; (b) graphs and especially 3D graphics are too difficult or abstract to be understood and handled by the common ur; or (c) the required technology is difficult to deploy, e.g. it requires proprietary browr extensions, such as Flas
h or Silverlight. In fact, veral
implementations of social graph visualizations exist as apps or social widgets bad on Flash. The u of Flash, however, restricts the u of such visualizations primarily to the desktop, as support for Flash is not common on mobile devices ( the discussion Apple vs. Adobe in 2010 (Copeland 2010)). Mobile devices on the other side are at the heart-beat of any SNP, as at the time being the networks live primarily from their immediacy.
We address (a) and (b) by prenting results of a small ur study, where we tested a basic ur interface using a 3D social graph, showing that urs prefer the graph version over the list-bad HTML interface in certain tasks. Urs are also generally faster when using the graph. Motivated by the results, we focus on the development of an interactive 3D social graph for Facebook, in which we put forward the idea of using emerging technologies, such as WebGL defined by the Khronos Group (Khronos 2009) and X3DOM, a DOM-bad integration model for HTML5/X3D (Fraunhofer IGD 2010b). The technologies will be part of tomorrow’s browrs and do not require proprietary extensions.
2. RELATED WORK
The visualization of social network data as graphs has a long tradition in the field of social network studies in social sciences. Freeman (Freeman 2000) reviews the evolution of techniques for visualizing social network data as graphs, starting as early as the beginning of the 20th century. Shape, color, position and the size of nodes can be identified as the main features for prenting distinct views on the data. However, the prentation of still images of a network is not always suitable if one has to extract meaningful data out of a network, especially with today’s social networks being very complex with a lot of different attributes connecting the individual nodes. Perer and Shneiderman (2006) stress the importance of adequate mechanisms for interactive exploration of social network data. They prent a rank-by-feature framework that enables network analysts to systematically detect certain patterns in the structure of a network. In (Perer & Shneiderman 2009) they prent their findings of veral ca studies evaluating their SocialAction framework. SocialAction provides a combination of visualizations and a toolkit for social network analysis mainly designed for rearchers. They conclude that rearchers can benefit from tools that integrate visualization and tools for exploration of networks.
While tools for scientific usage are common and rearchers are provided with powerful tools for network analysis there is still a lack of sufficient tools for end-urs of social networks to gain a deep
er insight in the structures underlying the data they are providing to the social network. Heer and Boyd try to address this problem with their Vizster tool (Heer & Boyd 2005) (Fig.1). Vizster combines the visualization of the social network with tools for exploration designed for end-urs. They evaluated Vizster in two ttings concluding that the possibility of the interactive exploration encourages urs to playfully examine their networks in more detail.
洪水之后In (Spritzer & Freitas 2008), a physics-bad method of interactively manipulating networks is prented.
晋江市教师进修学校
公主简笔画图片大全
A magnet metaphor is ud in the MagnetViz tool to provide urs with means to cluster nodes bad on世界十大珠宝
Figure 1: The standalone application Vizster allows end-
urs to interactively explore social networks in 2D. Figure 2: The Facebook app SocialGraph highlights clusters of friends in 2D as a single feature.
certain attributes. In (Bluhm et al. 2009) we prented a similar tool for the interactive exploration of
Last.fm in 3D, both on the desktop and in an immersive Virtual Reality installation.
A disadvantage of the prented tools is that they are standalone applications which are not integrated into the SNP. One of the rare examples of tools that embed into SNPs is SocialGraph (Phillips et al. 2010), a Flash application on Facebook, which us a force-bad layout to arrange personal friends in a 2D plane (Fig. 2). The single special feature of SocialGraph is the identification of clusters of friends, which are highlighted by circles. Thus SocialGraph provides a unique benefit over the web interface, as it allows the urs to clearly identify the social groups they are a part of. Interaction is nevertheless restricted to moving around nodes.
Until now there were only limited possibilities for the u of 3D graphics in browrs which might be a reason for the lack of 3D visualizations of social networks in the world wide web. In (Behr et al. 2009) an open source framework is prented that builds on top of WebGL, therefore no additional plugins are needed in recent browrs as is still the ca with Flash or Java among others. In contrast to WebGLU (DeLillo 2009), X3DOM is capable of directly integrating declarative X3D scenes into the HTML5 DOM tree, enabling the u of advanced JavaScript technologies, such as Ajax or JQuery, to interact with the 3D content. For a more detailed discussion of WebGL frameworks refer to (Behr et al. 2010).
3.GRAPH-BASED USER INTERFACES FOR SOCIAL NETWORKS
When advocating the u of 3D graph interfaces for SNPs, at least two issues are crucial: (a) the interface should actually help the ur, or at least be more fun to u, and (b) the interface should be easy to u. Both issues are addresd by recapitulating the results of a ur study done for an interactive graph-bad 3D desktop-browr called SONAR for the SNP Last.fm (Bluhm et al. 2009). The original paper is published in German; hence we give a conclusive summary.
3.1. Evaluation of a 3D Social Graph as a Ur Interface
Previously, we prented SONAR, Social Networks in Virtual Reality, an interactive 3D graph visualization of the enriched social network defined by Last.fm (Bluhm et al. 2009). SONAR is targeted at desktops and immersive Virtual Reality installations. Its implementation is bad on X3D and the additional functionality for immersive t-ups provided by the instantreality framework (Fraunhofer IGD 2010a). The tool allows the ur to literally immer into the social network and interactively explore the relationships between urs, their favorite music and artists, or similar artists. This allows urs to get a deeper understanding of their relationships and to identify patterns, which in the ca of the music network Last.fm might lead to the discovering of new interesting artist
s. We ud this application as tested to e, whether a 3D interface for social graphs can be understood by novice urs and if such an interface can provide advantages over the text-bad web interface.
In a ur study, 16 participants between 20 and 30 years, students at Bielefeld University and active urs of SNPs, were asked to perform veral typical tasks, such as exploration or arch, using the native web-bad frontend of Last.fm and SONAR (Bluhm et al. 2009). Both applications were tested on the desktop. Typical tasks were like (1) "Is A a friend of B?", targeting at direct relationships that are typically prented on a single page, (2) "Ur A likes song B, find a similar song.", which requires the ur to do a straight exploration into the depth of the network (that is, following up to two page transitions on the web frontend), or (3) "Find three mutual friends of urs A and B.", requiring the urs to follow at least two relations in parallel. The tasks were designed to be of increasing complexity regarding the overview of the structure of the local network excerpt required to solve the tasks. It was expected that tasks of category (1) were in favor of the text-bad interface whereas tasks of category (3) were clearly in favor of the graph-bad interface.
It turned out, that in all types of tasks urs were more efficient in terms of time-on-task using the 3D graph visualization of SONAR than with the web-bad frontend (Fig. 3). In addition to performance
measures, the urs were also asked which tool they preferred for each group of tasks. For tasks like (1), 69% preferred the web-bad frontend over SONAR, but for (2) 75% preferred SONAR and for tasks like (3) all participants preferred SONAR (Fig. 4). When asked directly, 83% of the urs also believed that they performed overall faster using SONAR.
快乐的新年
Figure 3: Time-on-task for the different task groups. All
tasks were procesd faster using the 3D graph. Figure 4: Preferences for the text-bad browr (grey) or the 3D graph (black). Urs preferred the browr
over SONAR only for the standard tasks (1).
The participants of the study were novices to 3D social graphs. Hence, the results for the 3D graph v
isualization in the study suggest that the 3D graph visualization can easily be understood and be ud purpofully, thus addressing issue (b). The fact that the 3D graph visualization is also preferred over the web-bad prentation for tasks were the structure of the social network is highly relevant also suggests that, (a), such a visualization is suited for typical tasks which are expected to be performed on a SNP.
3.2 Overt and Covert Structures of Facebook's Social Graph骨盆肌
Friendship is the constructing relationship of the popular Facebook social network. Urs create and manage individual profiles on Facebook. They can express their relationship to other urs by linking their profiles, which is a reliable declaration as both participants have to agree upon the relation. This friendship-relation is rather static, once accepted it would not change until the relation is revoked. Note also, that the friendship relation not necessarily coincides with a friendship in the real world. It could also express family relations, relations between co-workers or business contacts, among others. However, they are at least two sources of information made available by Facebook which are more conclusive about the relationships: communication events and evidences of real-life encounters .
探索创新Besides the structure-giving friendship-relation, Facebook supports communication events which can be interpreted as relations. For example, urs can directly exchange messages or they can upload images (authorship-relation), identify and tag other urs on photos or comment images. Comments are a universal tool provided by Facebook, nearly everything can be commented by other urs, be it messages, entries on the pinboard, photos, events or comments themlves. As the friendship-relation offers only a weak qualification of the link between urs, the number and kind of social interactions which can be obrved on Facebook, might provide a more thorough picture of the actual relationship between individuals and groups. The mining of the real social network from the interactions is currently a hot rearch topic.
All of the communication events happen in the virtual environment constructed by the social networking platform. The urs not necessarily have met in real life. Another rvice offered by Facebook provides also an index for real-life encounters : urs can upload photos and identify other urs who are depicted on a photo by tagging them. The prence of two urs on one or more photos constructs a relation have-met-in-real-life , weighted by the number of occurrences. This relation, however, has to be considered with care. If there is no photo showing a specific pair of urs, this does tell nothing about the have-met-in-real-life relation. A cond way to identify whethe
r two urs have met in real-life is to consider the event-management pages provided by Facebook. Events, such as parties, can have individual pages on Facebook which can be ud, e.g. for organizational purpos. Among other things, urs can be invited to events and they can state whether they have participated at the event or not.
However, as stated earlier, there is a lack of convenient tools for urs to explore and make u of the overt and covert information to their full extend.
4. A 3D SOCIAL GRAPH APP FOR FACEBOOK
In the following, issue (c) of deploying rich 3D visualizations on the web is addresd. WebGL and X3DOM are recent technologies which promi a smooth integration between interactive 3D content, the HTML-DOM and JavaScript. This is demonstrated by highlighting the interaction between HTML/X3D, JavaScript and the Facebook rver on lected features of a Facebook App for 3D social graphs (Mattar & Pfeiffer 2010).
4.1 Communication Flow
Facebook does not host third-party applications on their rvers. Therefore an additional rver is ne
eded for applications. This makes it necessary to authenticate the ur with an API provided by Facebook. Once this has been done and the ur is properly authenticated, the ur agent requests an application from within Facebook, Facebook responds with its so called Facebook chrome (Facebook's ur interface) and an embedded iframe for the application. The application is delivered by the application host and rendered into the iframe by the ur agent. Subquent JavaScript API requests are conducted by the ur agent and handled by the Facebook rver. Figure 5 shows the communication flow between the ur agent and the rvers in detail.
Figure 5: Communication between the Ur Agent (Browr), the application host and the Facebook rver.
4.2 Integrating HTML5 and X3D: X3DOM
The visualization is realized bad on X3D (Web3D 2008) using the HTML-Profile (Behr et al. 2009). For the embedding of the 3D scene which holds the interactive social graph, the X3DOM integration model (Behr et al. 2010) is ud. Fraunhofer IGD provides a reference implementation bad on JavaScript (Fraunhofer IGD 2010b) under the MIT/GPL dual licen. The project is hosted for everyone to contribute. Using X3DOM, the 3D scene can be declared directly in the HTML-DOM tree (Fig.
7).
4.3 Constructing a Live Graph
The visualization of the social graph is constructed in real-time and asynchronous by retrieving and analyzing relevant data from the Facebook rvers. This is all done in JavaScript (Fig. 6). If, for example, a new friend is detected, the according visualization (Fig. 7) is constructed by directly manipulating the embedded DOM tree (Fig. 7).
The X3DOM implementation monitors the DOM for changes and mirrors them in the scenegraph which is internally constructed and hidden behind the high-level X3D API. This way, X3DOM can optimize the declarative specification of the scenegraph given by the programmer and create a scenegraph which is optimized for performance, tailored to the WebGL implementation or, if available, a special purpo X3D plugin.