Analysing Posts About the #Coronation on Mastodon

There is no doubt that the recent coronation of King Charles III was a massive event in the UK. Whether you’re in favour of the monarchy or not, there is no doubt that the coronation event is a big talking point. This was certainly the case on Mastodon – my preferred microblogging platform of choice.

By focusing on posts which mention the hashtag over a seven-day period (3rd May – 10th May) a total of 4,269 posts were made across 1,997 users, 396 instances and featuring a total of 2,295 alternative hashtags. As shown in the timeline plot below, activity peaked around 6th May – the day of the coronation.

Across all 1,997 users that posted about the coronation, around 16.52% of which originated the instance – the largest instance of them all. Other notable mentions include, and The rest are recored below.

Instance Count Share 330 16.52% 101 5.06% 86 4.31% 75 3.76% 62 3.1% 61 3.05% 53 2.65% 50 2.5% 33 1.65% 31 1.55% 30 1.5% 27 1.35% 21 1.05% 20 1.0% 19 0.95% 19 0.95% 18 0.9% 18 0.9% 17 0.85% 17 0.85% 17 0.85% 16 0.8% 15 0.75% 14 0.7% 14 0.7%
The top 25 instances used featured in the set

To explore this space further, I built a user-hashtag network to consider the role of other, co-occurring hashtags. Building user-hashtag networks is something I have done in many previous posts (e.g. [1], [2], [3]). As shown in the figure bellow, the network is a little messy but, in order to reduce the network down to a more manageable size, I focused on the top 25 hashtags and users.

Among all the hashtags, the is a significant presence of anti-monarchy hashtags – a few of which include and . The remainder are fairly genetic.

One of the benefits of using bipartite graphs is that we can project nodes based upon mutual connections (as shown in a previous post). This means that we can connect hashtags together if they have at least one mutual user in common. The results are shown below.

With a large enough dataset, the projected graph could be used to perform some very basic topic modelling to see if there are any themes within the content. In our case, given the size, all of the hashtags are grouped together except for . That one cracked me up.

If you have any creative ideas on how I could better use this data. Let me know what you think. I’m always open to suggestions.