Where do DataNinjas go for help against that unbeatable foe called Gephi? That’s right, they ask online and wait for other ninjas to come to the rescue! They then ask them to write a guest post on their award winning blog. I am Sir Simon of Gephi and I nobly answered just such a call-to-arms.
Aside from being Silke’s occasional tech support, I also own the complete Adventures of Tintin and an Iron Age spear, so I have all of the relevant Belgian/classicist credentials to be here (also I’m working on a PhD that involves crime and SNA or some such nonsense). For a slightly different approach to spaghetti monsters to the usual DataNinja stuff, I’m going to highlight how to construct networks from twitter and tell the tale of a trip to a pub with some digital classicists…
Twitter digital classicist network
After helping Silke cure about a 1000 errors that were occurring when data was imported into Gephi, she was kind enough to invite me along to a Digital Classicist lecture at King’s to see what sort of thing is going on in the DC field. There was a jolly interesting talk given by Leif Isaksen and Elton Barker about the Pelagios project and Recogito, and perhaps even more interesting to someone interested in spaghetti monsters, people were being encouraged to use #DigiClass to interact with the presentation.
Twitter digital classicist network: Nodexl
There are various tools which can be used to gather this data, my personal preference is Nodexl which can be downloaded for free (although you do need a copy of Excel to run it and a Twitter account is preferable). I used Nodexl to collect tweets which contained #DigiClass, this information can then be exported as a GraphML format file for use in Gephi. Mine looked like this:
Which I thought was rather pretty, with a Yifan-Hu layout, and five colours representing five detected components and three isolates (who bizarrely were using the hashtag but their tweets were about other things). Apparently though as pretty as this network is, it wasn’t good enough for some (SILKE!) and so I created another which shows the usernames. This can be easily done in Gephi as long as there are label names in the data, fortunately by default Twitter data from Nodexl contains this information.
Another feature of Nodexl which can be rather handy is the ability to search for a particular hashtag and then collect data on the exchanges between users who had used this hashtag. For example in the above image we can see that palaeofuturist is the centre of a large cluster, but when we look at the exchanges between all the users in the above graph we may find that in fact there are in fact more central people to the community.
The graph below was generated to show the interactions between people who had used #DigiClass. In Gephi I have formatted this to be a dual-circle layout with the users with the greatest in-degree outside of the circle.
As it turns out palaeofuturist has received the greatest number of interactions from people who used #DigiClass, the strongest single link (the thick blue line) is harutaseiro interacting with eleanor_robson.
Twitter digital classicist network part 2: pub talk
Having learnt all there it is possible to know about Pelagios in an hour or so I was forcibly dragged to the pub, the walk to which gave Silke and I a rather geeky idea, why not document the interactions between the DC pub group and produce network maps from that? So we did (because we are geeks)! Our original plan was to also include the number of drinks everyone had but as the evening progressed this became increasingly difficult to follow and frankly we couldn’t be bothered. The interactions were recorded at various time points throughout the evening which we (well me) used to create dynamic networks in Gephi, so that it is possible to view how the networks evolved across the evening. Of course, not all interactions are included just who was talking to whom when we remembered to collect the data. Ladies are coded red and the chaps are blue. I simply recorded my screen whilst playing the network in Gephi to obtain the video.
This was somewhat time consuming to do but relatively straight-forward; you simply need a column set to TimeInterval format in Gephi with the information laid out in the following way:
or for more complex interaction <[start1, end1]; [start2, end2]>.
We had this information for both the interactions and the people and so in the graph you can see how people joined and left the network, and the evolution of the conversations during the evening. Suitably relaxed, and having been scolded for Belgian teasing I headed home a little more enlightened about the Digital Classicist community.