The following is a visual analysis of the use of the political hashtag #libspill in the hours prior, during and after the leadership vote, which retained Australia’s 28th Prime Minister as the leader of the Liberal party. For those readers, like myself, who maintain only a general interest in Australian politics and do not participate in the daily public discourse facilitated by hashtags like #auspol, this handy translation of the leadership ‘spill’ into Game of Thrones terms is incredibly illuminating. Similarly enlightening are the visual representations of the algorithmic clusterings of Twitter discussions, generated by the open source plugin for MS Excel, NodeXL, which can be used to provide a simple, but powerful diagrammatic analysis of the relevant hashtag use.
The approach here is very small scale, and my research interest is in the use of NodeXL as an ‘off-the-shelf’ application that requires no programming, coding or specialist training. My view is that micro-public data, and its analysis and management, is an important digital literacy, or perhaps a ‘network literacy’, that should be at the disposal of every social media user. The set of competencies, encounters and experiences that make up digital and network literacies are an especially important part of the contemporary skill set that students pursuing a Media and Communication Studies degree need to be equipped with in order to contribute and participate successfully in relevant careers and interests. We are already seeing the use of NodeXL and other forms of networked data visualisation for political and media reporting, but for a much more comprehensive ‘big-data’ approach, however, I recommend the work of Axel Bruns, Jean Burgess and others.
Figure 1. is the visualisation of a sample of 2000 tweets captured in the 24 hours from Monday February 9th at 8am, covering the duration of Twitter hashtag use prior and post the leadership vote. Each ‘node’ in the graph is an individual user who tweeted the hashtag #libspill, or is a follower of a user tweeting with the hashtag. NodeXL plots an edge (a blue line in this case) between two nodes if there is a relation in the form of a follow, reply or mention, and the software generates a visual representation of the hashtag use and the larger context of its Twitter network activity in the form of the graph.
The graph is prepared according to the methodology developed by Smith et al (2014), which identifies six clustering structures that are commonly observed in Twitter conversation and emerge because individuals selectively choose who to reply to and mention. Meaningful information can be determined from these graphs as they represent the expression of opinion, the citation of information sources, and the organisation of individuals into discrete micro-publics of follows and following, tweets, replies and mentions, and together these form the dynamic online conversation experience that is unique to the character limitations, tagging and other microblogging practices of Twitter.
Smith et al’s methodology is intended to be expanded on by drawing on further qualitative and quantitative approaches, such as surveys, focus groups, one-to-one interviews, and the data gathered by NodeXL can be used in sentiment, discourse and content analysis. Even in simple everyday use, however, NodeXL can provide an immediate way into the Twitter data that is not immediately obvious from the flow of Tweets that traverse our mobile and desktop screens:
“Our approach combines analysis of the size and structure of the network and its sub-groups with analysis of the words, hashtags and URLs people use. Each person who contributes to a Twitter conversation is located in a specific position in the web of relationships among all participants in the conversation. Some people occupy rare positions in the network that suggest that they have special importance and power in the conversation” (Smith et all 2014: 2).
The automated clustering algorithm options in NodeXL map the individual nodes according to the ways groups of users connect to one another, in this case placing people more connected to one another in different regions on the map. At first glance it appears that Figure 1. belongs to the Polarized Crowd network type, which is dominated by two dense and heavily oppositional groups. Polarized crowds are divisive, especially with regards to political topics and events, and they are characterised by very few connections between the groups, which indicates that members of different groups are not conversing, but ignoring one another and relying on alternative sources when discussing issues (Smith et al, 2014: 3).
A close look at the nodes in the large group on the right hand side of the graph shows the the primary group is made up of politically manifold personas; including former Australian Prime Minister, Julia Gillard, and Rupert Murdoch whose position in the graph is very close to Tony Abbott’s official Twitter handle, @TonyAbottMHR, which is presumably operated by a member of his staff given his recents comments of the value of social media as “electronic graffiti”.
Where the Polarized Crowds of network conversations indicates groups that are not connected by strong ties, the Tight Crowd network involves many connections between the dense networks of communities of Twitter users. The graph in Figure 1. is much closer to Tight Crowd structure in which individuals across the network are aware of each other and have conservations and exchange links and information. The large number of edges between the two dominant groups of users and the small number of isolates and less connected users in the lower right portion of the image, reveals the sharing of commons points of interest or significance, and a strong group of connections to others with similar interests.
The Tight Crowd networks in Figure 1. are “… composed of a few dense and densely interconnected groups where conversations sometime swirl around, involving different people at different times” (Smith et al 2014: 21). In the Tight Crowd network, argues Smith et al, there is no “other” group as is the case of the Polarized Crowd network. This is an encouraging view of Australian politics, which suggests more conversation, discussion and debate between the major political views of its Twitter users than is the case in the U.S. (see Himmelboim et al 2013).
NodeXl can be used to determine a number of important metrics from the Twitter data and meta-data of each tweet, including the most frequently linked to URLs and domains, hashtags, words, word pairs, replies to, mentions and Tweeters in the groups as shown in the following tables:
Top URLs in Tweet in Entire Graph
|Entire Graph Count|
|Top Domains in Tweet in Entire Graph||Entire Graph Count|
|Top Hashtags in Tweet in Entire Graph||Entire Graph Count|
|Top Words in Tweet in Entire Graph||Entire Graph Count|
|Top Word Pairs in Tweet in Entire Graph||Entire Graph Count|
|Top Replied-To in Entire Graph||Entire Graph Count|
|Top Mentioned in Entire Graph||Entire Graph Count|
|Top Tweeters in Entire Graph||Entire Graph Count|
Another method to expand the data collection and analsysis process is the use of commercial web-based services to collect and visualise Tweets. These sites vary in cost and sophistication, but I’ve found TweetArchivist to be a reliable and useful service to record every tweet/hashtag mention from keywords. The site provides simple but effective visuals, and the data can be exported as CSV files or PDF for later content analysis and further network visualisation with applications like Gephi.
The data collection from the archiving process included 22,624 tweets registering 109,013,514 impressions from February 9 8am to February 10 8am, 2015, and can be used to get a sense of the most frequent hashtag users, the distribution of the #libspill hashtag use in conversation in terms of the volume of Tweets over time, and using a range of factors including number of tweets, followers, retweets and replies to, we can review a measure of the ‘influence’ of Twitter accounts involved.
This is highly useful for those interested in #auspol and shows those media outlets with an active Twitter persona, and more easily observe the mixture of print and broadcast television news and entertainment organisations actively using Twitter. NodeXL makes it easier to dig further into this data and there is masses of detail to unpack and consider from these images and information. In the next update I plan to take a brief look at the use of the #ImstickingwithTony hashtag and a closer look at the role of #auspol in mediation of Australian political conversations.