14.1 Introduction

A wiki is a website anyone can edit, where every page modification is recorded and archived. The first wiki system, the WikiWikiWeb, was invented by Ward Cunningham in 1995 to allow a group to easily and quickly (“wiki” means “quick” in Hawaiian) edit a set of web pages without having to know HTML or deal with moving files back and forth to a web server. In addition to reducing the technical barriers to creating web pages, wikis make it easy for people to collaborate on writing tasks because the technology of the wiki provides separate spaces for people to create content and to discuss issues related to the content they are creating. Because of their flexible structure, support for discussion, and ease of use, wikis are an important platform for supporting online communities. For example, Wikia, a company that hosts wiki communities, has approximately 400,000 separate communities, and it is only one of many such platforms. This chapter explores how to use the activity within wikis to create social network representations that can help community designers understand who and what is important to a given community, determine whether the community is healthy, and distinguish between different types of contributors.

Although invented in 1995, wikis remained relatively unknown until about 2003, when Wikipedia, the online encyclopedia anyone can edit, started to come to prominence. Wikipedia has become the dominant source for encyclopedic information online, and it is increasingly understood as a social force that challenges traditional notions of authority, expertise, and knowledge construction [1,2]. Many organizations and educational institutions such as PBWiki, WikiSpaces, and Wikia, now use wikis as knowledge repositories for users. Open source projects and technical question-and-answer (Q&A) communities use wikis for documentation and support. Although most social media tools support information sharing, wikis are unique in their ability to support collaborative content creation and maintenance.

Wiki systems are also excellent sources of data for social network analysis [3,4], which can reveal important lessons about how people work together, what factors encourage long-term substantial contributions, and how leaders emerge. Alongside their richness, they are one of the more technically demanding forms of social media networks to extract and work with. Many wikis give rise to large datasets, which must be either parsed or sampled before any analysis can begin. Wikis offer many types of pages and several ways for editors to interact, providing a more complex social context than systems like email, chat, or Twitter. In this respect, wikis occupy a similar social and practical space as social networking sites like Facebook or Twitter.

The majority of wiki communities create information repositories about a specific topic. Whereas some wikis like Gardenology or the Marvel Comics Database take the form of an encyclopedia, other wikis encourage many genres of pages such as lists of annotated links, debates, or how-to documents [5]. The flexibility of wikis makes them highly adaptable to the needs of their users. Company employees use internal wikis to share information and coordinate effort, teachers use the wikis to collaboratively create and share lesson plans, cancer patients share information about clinical trials and doctors, intelligence analysts use Intellipedia to share information across agencies, fans of Star Wars use Wookieepedia to make sense of a complex universe, and Minecraft players use Minecraft Wiki to document the game universe and provide design resources.

Despite their wide use and potential as data sources, wiki systems constitute a challenging and complex form of online community with numerous modes of interaction and potentially dozens of ways to represent relations in the form of a network graph. Before beginning an investigation of any wiki system, it’s important to address several questions. Which modes of interaction are most valuable for understanding your online community? Which definitions of an edge are consistent with socially meaningful interactions in the system? What time frames of interaction should be used to demarcate the boundaries of your data collection? Because of these questions and because of the technical challenges associated with collecting many kinds of network data, wiki systems should be considered an advanced social media system for both practitioners and researchers. However, careful planning and data management techniques can make wikis one of the most rewarding areas for network analysis of social media.

The focus of this chapter is on introducing several exploratory methods for working with data from wikis in a social network analysis framework. There are many types of wiki networks that might be of interest for social network analysis. This chapter begins with a discussion of these network types and illustrates several questions that could be asked of wiki network data. It also describes attributes of wiki systems and addresses how those attributes may reflect the ways in which people interact in wikis.

The core of the chapter introduces the use of NodeXL and its built-in data MediaWiki importer to analyze the hyperlink and communication networks of pages across different languages in Wikipedia. While participation in the Wikipedia community allows users to connect with one another, talk, and share edits across geographic and political boundaries, the presence of language-specific Wikipedia communities means that cultural boundaries do exist within this space. Due to the crowdsourced nature of Wikipedia, these boundaries may generate differences in the content and category structure of pages addressing the same topics. The manner in which pages are framed, discussed and linked may vary according to the sociopolitical or cultural context of the author(s), and these identities may be expressed in part through language. Existing research has documented differences in how local and global historical figures are ranked by importance across 24 Wikipedia language editions [6], as well as documenting how language groups can map onto cultural differences across a wide range of different language wikipedias [7]. Similar patterns may be found in pages that address other cultural institutions, from popular fast food brands to political ideologies.

Data for this analysis were gathered using NodeXL’s MediaWiki importer. This tool is designed to download one of two types of primary relationships on Wikipedia—revisions and page hyperlinks—and allow researchers to build networks based on these associations. This chapter focuses primarily on page hyperlinks. On Wikipedia, hyperlinks between pages are created to link topics mentioned within the body of an article. For instance, a Wikipedia page on the social networking site Facebook mentions Facebook founder Mark Zuckerberg, and users are able to click on Mark Zuckerberg’s name to directly access his page. Hyperlinks between pages provide an emergent measure of relevance in the eyes of the community of active editors for a given article.

This chapter looks for differences in the page-to-page link structure for environmental and resource allocation pages across languages, and anticipates that these differences are attributable to the extent that the culture associated with the language is dependent on the natural resource addressed. Specifically, this chapter will focus on pages addressing whaling across Norwegian, Italian, and English language communities, and will note whether there exist differences in the content and structure of associated topics. Differences in the pages associated with each topic represent differences in each culture’s understanding of the topic. Institutions and concepts central to one language community’s network may not be present in the other.

14.2 Key features of wiki systems

Your tour begins with some of the key features of wiki systems that support studying the ways people behave in them. Wikis have a more complex structure than most of the other social media network data sources discussed in this book, such as Twitter or email. In particular, wikis keep a history of all editing activity that can be viewed on a per-page or per-contributor basis. They use namespaces to organize different kinds of contributions to the wiki. Wikis are built on the premise that everything is a page: the content itself, discussion of the content, individual contributors, supporting tools like category pages, and even the policies of the community itself are all contributed by users and evolve over time. Wikis also have user accounts, which track activity by each person, allow self-disclosure, and facilitate direct communication between people. This discussion will be based primarily on Wikipedia and the MediaWiki software that implements it; however, many wiki sites use the MediaWiki software, and the everything-is-a-page structure and namespaces that characterize Wikipedia are common to most wikis.

Most people’s view of wikis comes from seeing Wikipedia pages that result from a web search. For example, as of this writing, searching for “whale” on Google returns the Wikipedia page for Whale as the first result (Figure 14.1). The obvious feature of this page is that it presents information about the page’s topic. In Wikipedia, this generally takes the form of an encyclopedia article, although as mentioned in the introduction, these pages can provide other forms of information such as lists of links or instructions.

History. The visible topic page for any article is just a hint of the activity that happened in the course of creating it. In fact, every time someone edits a Wikipedia page, using the Edit source tab (note that this will show up as View source if you are not logged in, as in Figure 14.1), the software remembers who made the edit, why they said they did so, when it happened, and what changed. Clicking the View history tab within an article (Figure 14.2) will show a timeline of every change to a page. These edits can be browsed sequentially and are searchable by content and user. This historical information has allowed researchers to study everything from the overall growth of wiki systems [6] to effective patterns of collaboration in these communities [8,9]; it also supports the development of technical tools that make recommendations based on the topics in which people show interest [10]. You have a number of tools and resources for getting the history of a Wikipedia page or set of pages. Open source wiki software platform MediaWiki offers Application Programming Interface (API) endpoints for multiple wikipedia language communities. Wikipedia histories may also be accessed by external scraping tools. Data are also available via database dumps.

Namespaces. Namespaces organize different kinds of contributions. Sometimes, editors disagree on exactly what should be on a given page. Rather than continually editing each other’s versions of a page (a phenomenon known as an “edit war”) or marking up the page itself with comments and disagreements (which often happens, for example, when people collaboratively edit a Word document), the MediaWiki software separates article creation from article discussion using namespaces. The articles themselves are in the “main” namespace; for each article, there is an associated page in the “talk” namespace where people can talk about the page, ask questions, request clarification, and resolve disagreements without affecting the main page too drastically. For the main content pages of the wiki, clicking on the Talk tab (Figure 14.1) will open the related talk page. Although the underlying tool is the same, these talk pages typically take the form of threaded discussions. Figure 14.3 illustrates how contributors use talk pages to discuss page-specific edit decisions and to communicate normative standards of appropriate editing. Separating discussion from other editing makes it easier to study the specific ways people collaborate and the strategies they use to resolve disagreements [11–13]. The threaded nature of the discussions provides a natural way to create a social network based on Wikipedia activity, by creating ties between people who reply to each other in a discussion [4,14–16] (Chapter 10). However, capturing the reply structure from talk pages can be challenging.

Everything is a page. There are many other namespaces as well, including the File namespace, typically used for uploading images, sound files, or videos that are included in a page. The Category and Template namespaces, where contributors can create technical tools that group pages and create reusable snippets that can support everything from providing a standard set of information to be included for a category to welcoming new users quickly; and the Wikipedia namespace, where people create and debate Wikipedia policies [17]. These namespaces further organize and identify the kinds of contributions people make, allowing community designers to understand and support the roles people take on [15,16,18]. The fact that every type of wiki page is treated just like any other wiki page with a history of changes and edits supports the study of the evolution of the community itself. By analyzing networks of communication recorded via these edits, researchers can assess how individuals take on specific roles based on the kinds of contributions they make [15,18] and how the group makes policy and governance decisions [19].

User accounts. Although many wikis allow anonymous access and editing in an effort to reduce the cost of contribution, most regular contributors to wikis create user accounts. The same wiki feature that records the history of edits and changes allows researchers to examine the evolution of a page or study the edit activity of a specific person. Figure 14.4 illustrates a small section of a contributor’s edit log. The User namespace allows registered users to create a page describing themselves, their interests, their skills, and the parts of the wiki they are most involved in, which is a rich resource for understanding the characteristics of individual community members. The User talk namespace allows for direct user-to-user communication, another natural resource for creating databases representing social network relationships from Wikipedia activity.

14.3 Wiki networks from edit activity

There are many interesting ways to analyze Wikipedia based on the history of page edits and interaction of its users. Using social networking tools to analyze the structure of MediaWiki-facilitated networks, however, requires translating editor interactions and page-to-page connections into relationships with predefined entities and ties. Some of these networks are social and require us to operationalize what a tie between users signifies. In Wikipedia, a strong tie between editors is likely when two Wikipedians have shared edits on the same articles, reply to each other on article talk pages, and make edits to the user talk page of the other. Such strong relationships may signify an important collaborative relationship, or it may also signify an ongoing disagreement or dispute. The norms related to documenting edits, as well as the rules of respect and civility, result in qualitative documentation of the meaning of such interactions. For this reason, analysis of interaction edges assumed from edits should be fact checked against the qualitative record.

The network of hyperlinks embedded in an article may also be considered an ego-centric network, albeit one that represents concepts and topics related to an article. The inclusion or exclusion of particular concepts from any article is subject to rules of collaboration and contribution that are fundamental to Wikipedia, summarized by the “Five Pillars,” which follow: Wikipedia is an encyclopedia; Wikipedia is written from a neutral point of view; Wikipedia is free content that anyone can use, edit, and distribute; Wikipedia’s editors should treat each other with respect and civility; and Wikipedia has no firm rules [20] Because the emergent definition of relevance for a topic or issue stems from the Wikipedians involved in editing that article, the manner in which this topic or issue is embedded within a network of other pages likely reflects Wikipedian’s understanding of its content and significance. Though not explicitly social, similar metrics can be used to unpack the “relationships” within this content network. Hyperlinks between pages may indicate a strong conceptual association.

Networks are composed of vertices, or entities, that are connected through edges that represent the relationships between them. Both vertices and relationships can have attributes, such as the strength of a tie between vertices, or the length of time a vertex has been part of the network. As highlighted by Carter Butts [21], the challenge for the analyst is to choose vertices, relationships, and attributes that give insight into questions that matter. This section discusses each of these issues in general in the context of wiki systems and gives a number of possible example networks one might choose to construct.

What is a vertex? In many social network studies, vertices represent individual people or groups, companies, or institutions. In a wiki, each user account is a possible vertex for networks based on relationships between wiki system users. These vertices may be analyzed within the context of talk namespaces. However, for some questions, pages, and even categories of pages might be the appropriate choice for the vertices in a social network. For example, if you want to understand connections between the kinds of topics (or products) a community is interested in, making pages or product categories the star of the show is probably the right choice. The examples in this chapter use pages, as well as usernames as vertices.

What counts as an edge? If vertices are defined as users who contribute edits to a wiki, then edges may be defined as one of many activities that display some type of interaction between two users. Although there are many types of potential indicators of relationships between users, let us illustrate three common ways to infer edges between users from edits. First, edits from one user on the user talk page of another may be evidence of a directed communication relationship between the two of them. Second, edits on an article talk page may indicate directed “reply” connections. When editor A makes an edit that refers to a prior edit by editor B, it creates an edge from A to B, which creates part of the reply graph (Chapter 10). Third, edits of a content page may indicate an undirected connection indicating a shared interest. It may also indicate social interaction; in fact, editors who edit a page at about the same time often later go on to interact more directly [22,23]. For example, if editor A and editor B have both edited 4 of the same pages, you can create an undirected edge connecting A and B together with and edge weight of 4. This is a type of “affiliation network” where people are connected to one another based on their shared “affiliations” (i.e., pages they have both edited). Edges may also represent hyperlinks between pages (i.e., content). Articles referenced within another Wikipedia page create a directed edge representing a claim of relevance or connection. This network is similar to the network of hyperlinked pages that makes up the World Wide Web, though it is constrained to just the wiki pages. Another type of undirected link between pages can occur when multiple pages are “affiliated” with the same category. For example, if Page A and Page B are both classified under Category 1, Category 2, and Category 3, you could create an edge between the pages with an edge weight of 3. Finally, pages can be “affiliated” with one another based on having the same shared editors. For example, if Page S and Page T are both edited by 5 different people, an undirected edge with an edge weight of 5 would connect them.

What attributes are important? Researchers can leverage a variety of content to assign attributes to edges and vertices. A distinguishing feature of a user vertex may be the particular pages that they edit or abstain from editing. Edit activity may also be used to describe user vertices. The content and complexity of user edits can be measured, and the number of major or minor edits a user makes to a page may be indicative of their role within the wiki [15,23]. For edges between user vertices, the timing of an author’s edits relative to the edits of other users may be important. Edits indicative of communication on talk pages are likely to be in close temporal proximity, or to exhibit a structure known as the third turn, where an edit by one editor occurs between edits of another. Using edges to track the adoption of specific editing tools may also lend insight into which users are influential within their network [4]. Page vertices themselves have attributes that can be coded as variables: for instance, the topics or categories under which they fall, and the presence or absence of key hyperlinks.

14.4 Using the NodeXL MediaWiki page network importer to access Wikipedia networks

One key feature of NodeXL is its ability to download network data directly from wikis that use the MediaWiki software, including Wikipedia. Based on API tools available through MediaWiki, NodeXL Pro provides options to download relationships between networks and editors centered on a specific page within any MediaWiki domain. To access the MediaWiki importer, select the Import tab and scroll to From MediaWiki Page Network. This will open the dialog box shown in Figure 14.5. Given a seed article (e.g., Social_media) and wiki domain (e.g., en.wikipedia.org), this importer can create several types of networks shown in the Network portion of the importer.

As described in Section 14.3.1, there are a variety of networks that may be constructed via links and edits within Wikipedia namespaces, and NodeXL integrates an importer for collecting a range of networks from MediWikis [24]. Some of these networks leverage users as vertices, some leverage pages as vertices, and some display links between both. Edges may represent replies, edits or hyperlinks. Selections for network imports include:

• • User-User Network: Coauthorship This option downloads a specified number of revisions for a seed article (default is 50), finds all users who contributed these edits, identifies which pages the authors have edited, and generates an edge between users if they have edited the same page.
• • User-User Network: Discussion This option downloads a specified number of revisions for the talk page associated with the seed article, and generates an edge between users if their comments are sequential.
• • User-User Network: Article trajectory This option downloads a specified number of revisions for the seed article and generates an edge between users if they generate edits consecutively.
• • User-Article Network: Hyperlink Coauthorship This tool selects articles to which the seed article links and downloads a specified number of revisions for all of these articles (default is 50). For each revision, it establishes an edge between the user who generated the revision and the article in which the revision appears.
• • User-Article Network: Category Coauthorship Similar to user-article networks for hyperlink coauthorship, this tool downloads articles that are in the same category as the seed article, collects a designated number of revisions, and generates an edge between the user who generated the revision and the article in which the revision appears.
• • Article-Article Network This selection generates networks of articles and hyperlinks between them. It may capture hyperlinked pages that are 1.5 to two degrees separated from the seed article, and generates edges between them.

In addition to gathering networks, the MediaWiki allows researchers to specify desired attributes for vertices well. For instance, the resulting network may include data on the gender of the wiki user or the date they joined the Wikipedia community. These factors may be important to consider when analyzing dynamics such as network centrality or user influence. Additionally, you may limit the number of recent revisions downloaded, or limit to all revisions that occurred after a specified date.

14.5 Understanding topics through page-to-page connections

Our first example of wiki network research highlights the MediaWiki importer tool as it applies to hyperlink connections within wiki pages. This example illustrates how hyperlink ties between wiki pages can reveal differences in the development of those wikis and in differences in Wikipedians’ definitions of what is relevant to the discussion of a particular topic. It focuses on an environmental topic that varies in its current and historical significance—whaling—and explores how content linked to this page varies across Wikipedia language communities.

This example uses “whaling” as the seedpage and analyzes this within the English, Norwegian (Nynorsk), and Italian wikipedia communities. Whaling has long been recognized as an environmental and economic issue in English speaking countries. While some indigenous communities within English speaking countries still rely on whaling to harvest natural resources, this practice is widely outlawed for the sake of conservation. Whaling is of strong historic and contemporary relevance for Norwegian speakers. Despite global controversy, commercial operations within Norway continue to hunt whales for food and other products. While fishing and maritime activity is important to the economy of Italy, whales were historically a relatively unimportant resource and whaling is not practiced. Pages within language-specific Wikipedia communities are edited according to a standard set of Wikipedia-wide policies and norms, of which the “relevance” policy shapes what to include on a page and what to exclude. Therefore, differences in which pages are linked from the seedpage for an environmental issue will partly reflect the cultural and socially embedded understandings of that issue and related concepts.

In addition to the seed page “whaling,” this example also explores “International Whaling Commission.” The International Whaling Commission (IWC) was established under the International Convention for the Regulation of Whaling (ICRW) in Washington, DC, on 2nd December 1946. The preamble to the Convention states that its objective is to provide for the proper conservation of whale stocks and systematic advancement of the whaling industry. Their extended intention is to ensure that “hunts are as humane as possible for the whale and as safe as possible for the hunters” [25]. The Commission also conducts research, co-ordinates and funds conservation work on many species of cetacean. In 1982, IWC banned all commercial whaling by a moratorium which was objected by some countries like Norway. However, recognizing nutritional, cultural and aboriginal subsistence values of different parts of the world IWC allows controlled whaling in these areas. The pro-whaling nations like Japan, Norway and Iceland accuse the IWC of basing these decisions upon “political and emotional” factors rather than upon scientific knowledge–even its own Scientific Committee concluded that quotas on some species of whale would be sustainable. Given this controversy, the manner in which the three language communities selected address this organization will likely vary significantly.

14.5.1 Data collection and processing

This example uses data from three MediaWiki imports with “whaling” as the Seed Article in English, Norweigian (Hvalfangst) and Italian (caccia_alla_balena) within their respective Wiki Domains (en.wikipedia.org, no.wikipedia.org, and it.wikipedia.org, respectively). Each download used Article-Article Network as the desired network type with the network degree set to 1.5. This facilitates mapping articles connected to the seed article, as well as connections between these articles. The Download revisions from parameter uses the last date of the last documented revision. These search parameters are displayed in Table 14.1, using the Norwegian page as an example.

Table 14.1

Relevant Mediawiki importer selections for collecting the wikipedia hyperlink networks using a single page as a seed

Field	Value	Explanation
Seed Article	Hvalfangst	Whaling is
Wiki Domain	no.wikipedia.org	This is the larger of the two Norwegian Wikis (80% of the population uses this spelling system)
Article-Article Network	Checked	This defines the network edge as a directed hyperlink, where vertex 1 = page with link; vertex 2 = page that is linked to
Levels to include	1.5	Degree 1.5 networks reveal the structure of relationships among the vertices directly connected to the seed vertex
Download	1	Only one copy of the page is needed
Download revisions from	09/28/2018	Selecting the same collection date for all three pages helps ensure replication

After collecting these networks, it may be helpful to view only the connections between the articles directly linked to the seed page. Removing the seed page’s incident edges allows you to concentrate on the relevance structure among the degree 1.5 edges without distraction by the connection to the seed edge, which is redundant because it is implied by the import criteria. To accomplish this, navigate to the Edges worksheet and select the seed article in the visualized network. It should then be highlighted in the Edges worksheet, where you can set the Visibility to Skip. This will exclude the seed page from the network visualization and will make sure it is not included in graph metrics and clustering.

NodeXL’s Graph Metrics tool is used to calculate summary metrics (i.e., total edges, total vertices, vertex in-degree, vertex out-degree), degree centrality, vertex clustering and edge reciprocation (see Figure 14.6). It is important to note that hyperlink networks obtained differ in their level of development; networks are larger for more highly discussed topics. Therefore, simple network metrics like in-degree and out-degree will not necessarily be sufficient measures of centrality. However, both pagerank and eigenvector measures of centrality are scaled in a way that makes them comparable across networks. Within each degree 1.5 network, the pages with high eigenvector centrality (undirected) and high PageRank centrality (directed, in-degree) are most central. Due to different perceptions of whaling across communities, it is likely that concepts and issues described by pages that rank high in regard to eigenvector and PageRank centrality will be different within the Italian, English and Norwegian Wikipedia communities. Since this is a directed network, you can focus on those metrics most relevant to directed networks, though undirected network metrics can still be useful to examine.

14.5.2 Identifying key topics across Wikipedia language communities

The first step in this analysis involves generating networks that display all articles connected to the seed article (Figures 14.7–14.9). In this visualization, the Color and Size of vertices highlights the pages that were central in the discussion of whaling within each language context. Autofill Columns was used to set the Color of the vertices reflects their Eigenvector centrality (undirected centrality among central vertices). The orange vertices are least central, purple are mid ranked and the blue are most central when not considering the directionality of the network (i.e., assuming it is undirected). The Size of the vertices is based on their PageRank centrality (directed indegree centrality among central vertices), with the largest vertices having the highest PageRank. The colors and sizing used are within NodeXL’s default range because there were no extreme outliers in the data. Each network displayed uses the Fruchterman-Reingold layout, though the Harel-Koran Fast Multiscale would work well too by placing less central vertices along the edges and pulling the most interconnected toward the center. These visualization selections help illustrate attribute similarity across the three language networks, with some notable differences.

In the Italian language wiki article on whaling is framed as a geographic and historical topic, with some literary relevance. The majority of the more central pages refer to countries of the Atlantic such as Iceland, Greenland, France, and the Faroe Islands. Secondary and tertiary pages refer to particular centuries, years, and locations such as Nantucket, and the novel Moby Dick. While the page for the International Whaling Commission is included, it is not associated with other international bodies, treaties, or environmental groups. However, Japan [Giappone] (where whaling remains a contemporary issue) is connected to the IWC page and to Corea del Sur, though these are of secondary centrality. Notably absent are many scientific details such as particular species of whales.

The Norwegian wiki community appears to treat management of whaling in the international scene as a much more relevant issue than does the Italian wiki. The year 2008 is significant because important multilateral agreements were established, and the International whaling commission is central to the discussion as are nations with active contemporary participation in the scientific and international debate related to whaling: Japan, Norway, the United States, South Korea, in addition to the Faroe Islands, which played a major historical role in whaling practices. Secondary and tertiary terms include the whaling method of faeroysk grindadrap (a collaborative hunting method where multiple boats surround a whale and drive it toward shore), as well as many locations in the Faroes where whaling is currently authorized. The conversation related to whaling in the Norwegian wiki is not settled, many sections indicate that citations or references are needed.

The English language wikipedia defines whaling management and contestation as relevant, as well as scientific description of whale species, with the IWC, Japan, Norway, Greenland, whale watching, as central, along with a long list of whale species such as baleen, minke, bowhead, etc. Secondary and tertiary concepts include material (whale oil, harpoon), historical (Bangudae Petroglyphs), and social/political (anti-whaling, Sea Shepherd Conservation Society, United Nations). Similar to the Norwegian network, the International Whaling Commission is highly central, meaning that the English speaking Wikipedia community focuses strongly on whaling as an environmental and political issue. The larger number of vertices is likely also related to the fact that the English page is longer.

While these networks are thematically different, there is an interesting structural similarity across them. Investigating graph metrics reveals that the distribution of Eigenvector and Closeness centrality was bimodal across all cases. NodeXL’s Dynamic Filters tool illustrates this discontinuity and helps obtain a better view of the most central articles within each network by narrowing the visualization to include vertices with Eigenvector centrality values that fall just above the lower of the two modal peaks (see Figure 14.10). For the Norwegian network the minimum Eigenvector value was 0.0326, for the Italian network the minimum Eigenvector value was 0.0296, and for the English network the minimum Eigenvector value was 0.0186. The resulting networks, with vertices matching the Eigenvalue specifications, are displayed in Figure 14.11.

The high Eigenvalue subgraphs (Figure 14.11) indicate that the English and Norwegian Wikipedian communities both see the IWC as central to the discussion of whaling, while the Italian Wikipedians do not. While the English language is the most highly developed Wikipedia, the development of the Italian and Norwegian Wikipedias is similar to each other, and so the low centrality of the IWC should indicate a real difference in community held notions of relevance, rather than simply lower development of the wiki.

While reducing the network illustrations to include only vertices with high eigenvector centrality helps illustrate which pages are most influential in Wikipedians’ understanding of a topic, the presence of less influential pages may also be important. However, viewing the network structure as a whole makes it difficult to view smaller clusters. One way to view and compare these clusters is to generate a series of subgraph images. NodeXL’s Subgraph Images tool generates these automatically. Figure 14.12 displays generated subgraphs for the top 21 vertices in the language group datasets. Each subgraph is saved to a separate file, and files were sorted by image size as a proxy for subgraph degree.

Examining these sub-networks reveals which entities and topics are highly embedded within the page structure of these three language contexts, and the extent to which they are linked to broader conversations. They illustrate that international governing bodies are relatively important in both the English and Norwegian Wikis, but not in the Italian Wiki. Overall, the Norwegian Wiki focuses on whaling in the context of the global community (see: mention of the United States, Japan, South Korea) as it relates to environmentalism (see: mention of Greenpeace), whereas the English wiki addresses whaling as an international issue (see: mention of Greenland, Australia), as it relates to native communities and traditions (see: mention of Inuit, aboriginal). Italian Wikipedia also recognizes the international importance of whaling (see: mention of Greenland, Spain, Norway) and addresses the historical significance of the practice (see: mention of 14th century, medieval) but with less focus on native populations than the United States.

These graphs illustrate how NodeXL may be used to import data from Wikipedia and map networks of pages that lend key insight into the meaning and significance of a specific topic across cultural contexts. By examining the overall network structure and focusing on degree centrality of articles linked to “whaling” across the English, Norwegian and Italian Wikipedia communities, differences emerge. For Norway, one of three countries in which whaling is legal, the presence of vertices addressing the International Whaling Commission and specific whaling practices make clear the importance and controversy of global regulations. English speaking countries recognize the importance of global whaling regulations, but this topic is largely of historical and indigenous significance. Italy, which does not share Norway or North America's historical and/or current reliance on whaling as a commercial enterprise, is less involved in conversations addressing whaling regulations. Overall, the structures of these pages and the structure of cultural conversations on this topic parallel one another closely.

14.6 Analyzing the structure of discussion page interaction

In addition to analyzing the structure of page-page link networks, you may also use the MediaWiki importer to gather and analyze the user discussion networks underlying the creation of these pages. Given that there are clear differences in the meaning and cultural salience of whaling across language categories, there may be differences in the extent to which contributors across languages discuss article edits and manage disagreements.

14.6.1 Mapping networks and identifying disputes within the English International Whaling Commission talk page

As described in Section 14.4, NodeXL's MediaWiki importer can collect edit history data from article talk pages. The next example uses discussion networks related to the English Wikipedia page for the International Whaling Commission, which were collected using the MediaWiki importer. Set the Seed Page to International_Whaling_Commission and the Wiki Domain to https://en.wikipedia.org/wiki/Talk. Then select Discussion from the User-User Network to download all users associated with the sequence of comments on the talk page.

By default all wiki pages have the potential for a talk page to be established, but these articles lacked talk pages for both the Italian and Norwegian language groups. Because talk pages are important locations for Wikipedians to solve disputes, their absence may suggest that the contributors to the ICW pages in Norway and Italy lack disputes or substantial disagreement. In the Norwegian case, it may be that all contributors approach the issues from the same political side while the Italian contributors see the situation as a third party without a clear agenda. Thus, this analysis focuses only on the English talk page.

The current version of the MediaWiki importer for talk pages uses the order of edits on the history page to infer edges. Figure 14.13 displays a default network based on the user talk page for the English Wikipedia International Whaling Commission page. The Sugiyama layout provides a good starting place for rendering threaded discussions within this visualization, the color orange indicates the presence of the terms “dispute” or “allegation” in the talk page edit text, highlighting the dyads engaged in conversation related to controversial issues. These were identified using the Word and Word Pair Metrics dialog as shown in Figure 14.14. Specifically, the content in the Edit Comment field on the Edges worksheet was used and List 3 was set to include values with the terms “dispute” or “allegation” (see Figure 14.14 coding of word pairs). This layout (Figure 14.13) indicates that Wikipedia user SammytheSeal is a central player within the talk page, and that this user engages in relatively frequent two-way disputes with user matt77 and some directed disputes with user Jonathanmills. User Andrew Galvenize-Davies, a first-degree connection to SammytheSeal, has reached out to both UberScienceNerd and Istvan with disputes or allegations. It is possible that SammytheSeal and Andrew Galvenize-Davies are experts on the topic, and are sensitive to misinformation.

This network map is a good start and provides useful information about active reply structures. However, there are key elements that it neglects. For one, because each page is divided into sections and not all comments are threaded, sequential commenting does not necessarily indicate a reply. Related to this, the structure implies that many users reply to themselves. It is unlikely this is the case, and possible that they simply contributed to different segments of the talk page in sequence. Finally, this structure ignores the roles that bots play within Wikipedia. Sinebot, for instance, adds signatures to comments on talk pages. These bots do not “discuss” topics in the same way as other users. These networks must therefore be restructured to better illustrate which users directly communicate with one another, and how bots may nudge or direct their conversation (Figure 14.15).

Mapping a discussion network that more accurately reflects conversation structure involves cross-referencing the talk page and talk page revision history with the default MediaWiki output. To accommodate this change, two new vertices were added: “FellowWikipedians” for posts that initiate a new topical subsection in the talk page, and “Cleanup” which refers to edits to page attributes that are not direct contributions to the discussion (often changes in links, categories and other dependencies between this page and other parts of the wiki). The resulting network (see Figure 14.15) helps clarify important relationships within the talk page network. It separates page maintenance communication from page discussion (see ties directed toward vertex Cleanup), and allows you to view which users initiate discussions within the talk page (see ties directed toward the vertex FellowWikipedians). Within this network, it appears that user SammytheSeal still engages in disputes—often with user matt77. However, this user is not directly involved in the construction of the talk page because they do not initiate any discussion topics. Reading the content of SammytheSeal’s contributions shows that this user actively resolves disputes and seeks consensus through reference to relevant citations by engaging in deliberative discussion [19,26].

The revision of the network structure depicted in Figure 14.15 also draws attention to an important social role [15] among Wikipedians, the “wiki gnome.” Arctic gnome, Epipelagic, and others linked to cleanup are engaged in small, inobtrusive tasks that need to be done. Similar to some bots, these wikipedians spend some or almost all of their edits repairing links, updating categories, and similar tasks. While some tasks can be standardized enough to become the focus of bots, others require enough situational judgment that the wiki gnome role remains important.

14.6.2 Identifying productive members of a talk page community

Given that the MediaWiki importer also captures text describing interactions within the Edit Comment field, researchers may also incorporate content analysis into studies of talk page network activity. Studies like this offer great ways to combine qualitative analysis with social network visualization and present viable strategies for professionals and researchers interested in teamwork, small groups, and interpersonal interaction. The example provided in Section 14.6.1 highlighted ties that denote “dispute” or “allegation” using orange. Researchers interested in using the MediaWiki importer may also consider using this tool to analyze other conversation properties and user roles.

For example, visualizations can use content from the Edit Comment field to highlight which users are most productive within a user talk page. A productive user may be defined as one who frequently engages editorial tasks or initiates page discussions. Web administrators will want to encourage productive discussion to help avoid edit wars or other dynamics that can squelch widespread participation in the wiki. Identifying which users lead through productivity could help facilitate this. The following example uses user talk edits for the English Wikipedia article on whaling.

The first step in this analysis involves generating a network graph similar to what is displayed in Figure 14.16, in which edges are weighted according to the number of replies and vertices are sized according to each user’s centrality to the conversation. Productivity is measured by counting commonly observed word pairs related to editor leadership actions: “major rewrite,” “new section,” “section article,” and “article concise.” After identifying this content, the NodeXL Autofill Columns feature adds these counts to the graph (vertices with higher counts are more blue). The resulting graph helps researchers to visually identify users (like SammytheSeal, Enuja, and Pcb21) who are both active and productive within the construction of the article. Furthermore, it allows one to note whether there are structural similarities regarding where these users are within the broader page network.

14.7 Choosing the right sample frame for your wiki research

One challenge associated with conducting wiki research involves bounding your sampling frame. The sampling frame describes the set of edits that you chose to include in your sample and from which your units of analysis will be assembled. You will want to define a disciplined rationale for the inclusion of edits in your study. It is also wise to begin with a carefully selected but limited framework because it is easy to become swamped in too much data.

The analysis of discussion within the International Whaling Commission talk page included all edits because the page was not excessively large. If it had been, it may have been wise to select a particular temporal range of edits, or focus only on edits for topics with the most directly interactive conversations. Controversial topics may include multiple archives of talk pages, in which case the proper sample frame is likely to focus on a limited set of topics or section headers within the talk page.

For the discussion of the IWC talk page, the validity of collaborative role determinations can be improved by expanding the sample frame for key editors. The initial investigation suggested the SammytheSeal was playing a key collaborative role, while Istvan was advocating for a particular political position (as a member of a well known environmental group). An improved sample frame would investigate the edit histories of both Wikipedians, especially their other contributions to article talk pages to see if their pattern of collaboration indicates a consistent social role across multiple topics.

The analysis of page link structures used the idea that only those pages that warranted a direct link from the seed page (i.e., whaling article) were the concepts most directly relevant to the discussion. This sample frame can be improved by investigating related concepts such as seal hunting or overfishing, and making predictions about how these concepts are likely to intersect with the cultural boundaries related to the language groups.

In general, it is helpful for sample frames to be drawn narrowly and then expanded incrementally. Internal wikis often grow quite large, it is always best to start your data collection with clear, theoretically based justifications for inclusion of data. Too large a dataset may pose problems in terms of the time and resources needed to manipulate and analyze the network graph. Beginning with a manageable and well reasoned sample frame will help your research be interpretable and replicable.

14.8 Practitioner’s summary

This chapter illustrated several structural visualization strategies that can be readily applied to any type of wiki: examining the structure of page links as a means of illustrating understanding of a topic, and identifying which contributors communicate and engage in disputes within a particular page. Although practitioners with access to corporate wikis or other internal datasets may find it easy to formulate research questions, it will typically be much more difficult to identify the best path for answering them. This chapter addresses key steps in overcoming these challenges. First, familiarity with the wiki system in general will help practitioners identify which types of pages and which projects warrant special attention. Second, practitioners should use knowledge of their particular projects, particular populations of editors, or time periods of observation to narrow their attention. Clarifying which questions need to be answered will help practitioners identify which edits to use as evidence of meaningful network relationships and which attributes are relevant and should be measured. Both substantive and theoretical insights are crucial for narrowing the sampling frame in ways that will best answer research objectives.

For those engaging in exploratory work, NodeXL is a useful tool because it invites the researcher to be creative. Wiki network studies are a rapidly developing area of research. With just a computer and an Internet connection, your may uncover patterns among wiki collaborators or hyperlinked pages never seen before. The studies in this chapter just begin to scratch the surface of what can be done with wiki-edited network data. The building blocks of analysis and measurement throughout this volume can be combined and applied to wikis, and doing so will likely be fruitful.

Part of the impetus behind the development of the NodeXL project is to expand the range of researchers and practitioners who would consider using social network analysis techniques in their research. For many years, network analysis software exacted a high price from people wanting to use these tools in terms of technical skill and willingness to learn proprietary software systems. NodeXL as well as browser-based network visualization tools like TouchGraph are helping expand participation in social network analysis to those without programming skills. The accessibility of these tools opens new pathways for established research questions, and, most importantly, turns the lens of network analysis on new questions and subjects. Because wikis are rich in terms of content and interaction, the growth of new research agendas in the social network analysis of wikis will be especially fruitful.

14.9 Researcher’s agenda

The ability of NodeXL to increase accessibility of network analysis opens fascinating new pathways for research. As this chapter illustrates, the ability to link pages between wikis and generate a network of references provides valuable information regarding how wiki contributors—and their cultural context—interpret a particular issue. This chapter illustrates how pages related to whaling vary according to how dependent the cultural context is or has been on the resource. However, researchers may also consider assessing other topics that invite controversy—such as vaccination—and assess the extent to which conspiracy influence and/or partisanship is influential in this network. Such work may consider how the structure of pages changes over time, as conversations surrounding a topic evolve and new ideas and issues are added or removed.

Wiki data may also provide an unprecedented window into the process and product of distributed collaboration. The cooperation and coordination problems inherent in any large task are, to varying degrees, overcome in different wiki projects. How is this collaboration achieved [27,28]? What are the key conditions that shape differential levels of collaborative success [8]?

Online and off, people frequently adopt distinctive patterns of behavior and interaction known as social roles [15,29]. Wiki systems are no exception; in fact, because of the complexity of the system and diversity of the organizational tasks, wikis are likely to give rise to many distinctive social roles [16,18,23].

Finally, there are other possibilities for network analysis using wikis not addressed by these analyses. For instance, wikis are rich settings in which to study the dynamics of diffusion [4,30,31]. The temporal history of wiki edits is retained with high fidelity, and editors have both durable identities and numerous internal contexts for interaction; researchers can trace diffusion processes better in wikis than they can almost anywhere else. Bridging research on diffusion with studies of cooperation, as well as with those of social roles, will be a fruitful direction for further research.

The intention of this chapter is to encourage researchers to make greater use of social media data collected from wiki resources because they offer one of the best combinations of socially relevant, accessible and consistent sources of social media data. While some social media systems are short lived, and others may suddenly restrict access, wiki systems like those supported by the Wikimedia foundation offer a reliable social media source for the study of culture and collaboration.