Coverage of our report on the six basic types of social media network structures created with the Pew Internet Research Center has been extensive. Here is a round up of the articles we have found about the study.
Mapping Twitter Topic Networks:
From Polarized Crowds to Community Clusters
The paper documents the distinct patterns of connection that emerge when people talk to one another using social media services like Twitter. The paper includes six network visualizations that clearly demonstrate the diverse ways people connect to people when using online tools.
Ballroom ABNetworks are everywhere, particularly in social media. Understanding networks can quickly reveal the key people, groups, and topics that matter most. But the tools to collect, analyze, visualize, and gain insights into connected structures have remained complex. Now the free and open NodeXL application makes network analysis tasks as easy as making a pie chart. The Network Overview Discovery and Exploration add-in for Excel (2007, 2010, 2013) extends the familiar spreadsheet, enabling users to easily access networks from a range of data sources including Facebook, YouTube, Twitter, Flickr, email, message boards, Wikis, blogs, and other repositories of connections. With simple automation tools, NodeXL users can calculate a range of network metrics, create a visualization, and generate a report highlighting key people, groups, and top URLs, hashtags, words and word pairs used in the discussion network. Network maps have revealed many of the hidden structures of social media, defining the major differences in the shapes and structures created as people link to one another.
If you have questions on social network analysis, meet with Marc to talk about:
NodeXL and related network analysis and visualization tools
How to collect, store, analyze, visualize, summarize and publish social network reports with just a few clicks (and no coding)
How to identify key influential people and subgroups within a conversation network
How to apply social network analysis to social media marketing
How to apply organizational network analysis to enterprise collaboration
Above is a map of the connections among the people who recently tweeted the term “strataconf” over the 7-day, 19-hour, 38-minute period from Sunday, 26 January 2014 at 21:53 UTC to Monday, 03 February 2014 at 17:32 UTC. The key people in the network at this point are:
You can make these types of maps with just a few clicks using NodeXL.
This is a map of the network of 2,785 Twitter users whose recent tweets contained ““kansas state” OR KState” over the 1-day, 23-hour, 14-minute period from Monday, 13 January 2014 at 17:06 UTC to Wednesday, 15 January 2014 at 16:20 UTC.
Graduate students in Computer Science at the University of Maryland in a class on information visualization produced a striking variety of NodeXL network analysis visualizations for their recent homework projects. The class, taught by Prof. Ben Shneiderman (www.cs.umd.edu/~ben), covers commercial tools, such as Spotfire and Tableau, and research software, giving students a chance to learn a range of existing visualization techniques and tools. The NodeXL homework project is done by individual students, midway in the semester, while 5-person student teams are also busy working on their major term projects, which create novel visualization tools for specialized applications. To see all the projects, click:
(Don’t be deterred by security warnings, the class wiki is open for all to read, but only students can edit)!
Several of the 30 projects deal with Facebook, Twitter, email, Wikipedia, and YouTube social networks, with academic citation patterns and sports networks adding variety. Entertainment, finance and medical analyses round out the collection, showing the huge range of potential NodeXL applications. Students had only two weeks to find data, import it, clean it, and then create meaningful visualizations that enabled them to find interesting insights into connected structures.
Gregory Kramida’s analysis of stock symbol co-occurrences in financial articles
Gregory Kramida analyzed the connections among company names in the business press. See:
The project shows the strong linkages between technology companies and consumer services, finance and public utilities. The data set of more than 50,000 financial articles had more than 400,000 co-occurrences of stock ticker symbols. He used the NodeXL grouping feature to organize the stocks into groups by industry and then showed results using the Group-in-a-Box layout feature. This network is limited to companies that were mentioned together at least 50 times.
Ruofei Du’s analysis of co-authorship patterns
Ruofei Du probed the relationship among authors in 1033 scientific papers from the 1988 to 2013 User Interface Software & Technology (UIST) conference. See:
The co-author collaborations followed commonly seen patterns of professors and their students, but the relationships between academia and industry showed novel patterns. After grouping authors by their organizations, it is apparent that Microsoft is well-represented at this conference through numerous collaborations with universities.
Joshua Brule’s analysis of actor co-performance connections from the television series Firefly
Joshua Brule created an intriguing story of television and film actors and actresses that emerges from analysis of ten actors from the cancelled television series Firefly. See:
The actors had few collaborations before appearing on the program, but many afterwards. The carefully constructed bipartite network shows how ten actors collaborated in 38 films, television shows, or videogames.
Network analysis is a way of looking at the world that focuses on the shape and structure of collections of relationships.
In a network perspective the world is not primarily composed of individuals (“nodes”, “vertices”, “entities”). Instead, a network approach focuses on relationships between individuals (“edges”, “ties”, “connections”, “links”).
When collections of connections are analyzed, network patterns emerge. Networks have a variety of shapes and within them people occupy a variety of locations within each network. Some people are highly connected, while most people have just a few connections, for example.
Network theory provides a big collection of math that enables the measurement of these shapes and structures.
Using these measures, network analysis can identify key people in important locations in the network (for example: hubs, bridges, and islands). Network metrics allow for the network as a whole to be measured in terms of size and shape. Networks have many basic shapes and we have found six shapes to be common in internet and enterprise social media: divided, unified, fragmented, clustered, outward hub and spoke, inward hub and spoke. These shapes are created when people make individual decisions about who to reply to, link to, and like.
Divided networks are created when two groups of people talk about a controversial topic – but do not connect to people in the “other” group. Unified networks are formed by small to medium sized groups that are obscure or professional topics, conference hashtags are a good example. Fragmented networks have few connections among the people in them: these are often people talking about a brand or popular topic or event. Clusters sometimes grow among the people talking about a brand, indicating a existence of a brand “community”. Broadcast networks are formed when a prominent media person is widely repeated by many audience members, forming a hub-and-spoke pattern with the spokes pointed inward at the hub. The final pattern is the opposite, hub-and-spoke patterns with the hub linking out to a number of spokes. This pattern is generated by technical and customer support accounts like those for computer and airline companies. Additional patterns may exist, but these patterns are prominent in many social media network data sets.
When applied to external conversations, social media networks help identify the “mayor” of a hashtag or topic: these are the people at the center of the network. Network maps can be compared to the six basic types of networks to understand the nature of the topic community. We can look for examples of successful social media efforts and map those topic networks. Social media managers can contrast their topics with those of their aspirational targets and measure the difference between where they are and where they want to be.
When applied to enterprise conversations and connections, network analysis can reveal the experts who answer many people’s questions and “brokers” who bridge otherwise disconnected groups as well as the “structural holes” that show where a bridge or link is needed.
These insights can be useful in mergers, HR evaluation of group and manager performance, and identifying internal subject matter experts.
Research performed using NodeXL shows that work teams that have higher levels of internal connection (which is called “network density”) have higher levels of performance and profit. See:
The impact of intragroup social network topology on group performance: understanding intra-organizational knowledge transfer through a social capital framework Wise, Sean Evan (2013) The impact of intragroup social network topology on group performance: understanding intra-organizational knowledge transfer through a social capital framework. PhD thesis, University of Glasgow.
Full text available as: PDF Download (2499Kb) | Preview http://theses.gla.ac.uk/3793/