Sometimes a picture paints a thousands words, and that’s as true in data science as it is in anything else. In this post I want to capture a day (Sunday 12 April 2015) in the life of the General Election – namely the social graphs around key topics of the day.
The actual graphs themselves are huge and are not practical for a blog post, so I’ve taken screen captures of them. If any one is interested in seeing the full graphs, then contact me and I can arrange to make them available to you.
In the pictures, themselves, the edges (Twitter accounts) are too small to be seen, but the vertices (the lines between the edges) give an idea of the volume and direction of information flow. I also analysed the content of each tweet and colour each vertex with a party appropriate colour, to show which party is the subject of the the tweet.
Below is the full graph for the day:
You can see, at the top and the bottom of the graph, numerous accounts having little conversations amongst themselves, and in the middle, the majority of activity directed towards the main parties. Notice the dominance of the blue (Tory) and yellow (SNP) vertices, showing that conversations about the SNP and the Tories dominated the GE2015 stream, on that day. (The SNP generally dominate; whilst the Tory chancellor made a less that stellar appearance on a Sunday politics show).
Let’s go on to explore a few topics, firstly the #VoteSNP tag:
The #VoteSNP tag is at the centre of the round ball. Notice there are a good number of tweets with Labour as the subject hitting that tag. The edge on the right, with all the blue vertices heading into it, is the account of Jim Murphy, leader of Scottish Labour. I believe this reflects the attack on his party, by the SNP, as being “RedTories”.
By way of contrast, here is the #VoteLabour tag:
As you can see, in contrast to the #VoteSNP tag, the Labour tag is really an echo chamber, with Labour dominated subjects; this may seem intuitive, as these tweets are all tagged with #VoteLabour after all, but there is much less diversity of subject here than there is with the #VoteSNP tag.
Let’s look at the counter tag, #SNPOut:
The tag occupies the larger circle, in the centre of the image. We can see mainly SNP and Labour orientated vertices heading into it, as we would expect as Labour and the SNP are the two main parties in Scotland. The smaller clique, in the lower right, is centred around UKIP; it’s not immediately obvious what that would be about and it bears closer analysis.
Talking of UKIP, let’s take a look at their social graph:
As you would expect, #UKIP is the big edge in the middle. Surprisingly, as most people associate UKIP and the Tories as being similar, as they are both right of centre parties, there is not much blue to be seen on this graph, it is mainly purple (UKIP) and red (Labour) with a little smattering of the other parties.
Let’s look at a cross party issue, the #NHS:
This is, perhaps, the most interesting social graph. The first thing to notice is that it’s all blue and red, as the two “big beasts” of the election duke it out over the NHS.
Another interesting fact is that the graph is devoid of any yellow (SNP). The NHS is devolved in Scotland and so any failings therein lie squarely at the door of the SNP, consequently, they have not mentioned it much in this campaign.
#NHS occupies the large edge in the middle of the graph, the satellite edges are mainly related to Osborne’s, less than stellar, appearance on the Marr Show.
The SNP took a bit of a beating at the start of the week, and when that a happens it’s never long before the issue of independence raises it’s head, so let’s take a look at the social graph for the #IndyRef tag:
The #IndyRef tag is the edge on the left, with political views of all colours represented. The edge on the right is the #SNP tag (in relation to #IndyRef), it’s much more of an echo chamber.
Well that concludes this post, remember that these graphs were generated to give you a snap shot, in pictures, of a single day in the campaign, no inference other than that can and should be drawn from this data.
Until next time, keep crunching those number!