Continued on from this post. Also see my supervisor’s comments.
Interestingly, it turns out there is an Eclipse product called “Eclipse RT”, so perhaps there weren’t as many retweets as the wordle might have suggested…
Still – 413 of the 640 tweets in the data set contained an @ mention of some kind. Of these, 72 were “directed” (by which I mean, start with an @).
I created network graphs using Prefuse. There are two kinds of links – included in an directed tweet (yellow), and included in a non-directed tweet (purple).
This split is because to me there seems a clear distinction between a conversational directed tweet, and a mention – at a conference, if someone’s speaking a tweet might consist of <quote> + <credit> – this is sharing the person’s ideas with the audience of your twitter stream, but will not normally be denoted by retweet notation.
As yet, I haven’t created and uploaded an applet – for now, you can see the screenshots below. Some observations:
- There’s a dense graph, which the vast majority of users are not connected to.
- There are a few two-person connections separate from the graph.
- The bulk of the activity takes place in this core.
I think it’ll be interesting to apply some clique finding to this. I’m also interesting in applying this distinction (directed vs. mention) to temporal rhythms – thinking along the lines of what I created here.