Early Notes on Conference Tweeting

I just did a run on the first two days of the 2009 Nonprofit Technology Conference using the tools I've been working on (see my post on #swineflu earlier this week.) Using the hashtag #09ntc, I parsed 3834 tweets, and I looked up the hubs and authorities, plus generated the graph of the largest strongly connected component within the larger directed graph created from all the "@" replies and retweets.

Some really interesting stuff came up. Clay Shirky came up as the highest-ranked authority in the Twitter stream, based on the HITS algorithm. This makes sense--Shirky was the keynote speaker at the conference, and many people using the #09ntc hashtag referenced him. What's cool is that he doesn't show up, at all, in the hubs, something I've tended to see before in other conferences I've looked at. Looking through his recent tweets, he didn't use the #09ntc tag at all. He is, for the purposes of this conference, only a source of information, not a reporter of it. This is, of course, not to suggest that Shirky is a bad twitterer--he's anything but (I'm one of his now 15,648 followers). He just had a different function in the Twitter backchannel for #09ntc.

The highest ranked hubs are Steve MacLaughlin, Beth Kanter, and rachel ann yes. Of these, kanter and rachelannyes are also in the top of the authorities. While smaclaughlin is doing a lot of valuable reporting, kanter and rachelannyes are doing so as well AND are being replied to and retweeted more (from high-ranking hubs) within the channel. Not sure why that is--I'd be curious to hear from people who were there--leave a comment or hit me up on Twitter.

The subgraph representing the largest strongly connected component for #09ntc is huge. Enormous. You can download all 2.6Mb of the JPEG file, if you want to see for yourself. Of the half dozen or so conferences I've been looking at, this is, by far, the biggest SCC I've seen. It indicates, to me, that this crowd is doing a LOT of replying and retweeting with each other. My guess, and it's only that at this point, is that #09ntc is a very social group, as compared to other conferences that are largely broadcast events (this seems to be what I've found from the 2009 Game Developers Conference, which had many more tweets coming from it, but a much smaller SCC.)

Anyway, just another snapshot from me of some informal research I'm doing into both scheduled and spontaneous events. Hopefully, this will be useful to other folks--it's been really interesting to me so far. Here's some more of the raw data:

Betweenness centrality

andrewjcohen 0.012
webb 0.014
kalabird 0.015
rachelannyes 0.015
johnmerritt 0.017
silverbell 0.019
09ntc 0.026
smaclaughlin 0.027
kanter 0.034
ntenhross 0.036

Indegree centrality

nonprofit_tech 0.027
johnmerritt 0.029
chadnorman 0.029
silverbell 0.030
rachelannyes 0.030
kalabird 0.034
09ntc 0.046
ntenhross 0.071
kanter 0.091
cshirky 0.104

Outdegree centrality

rachelannyes 0.022
ntenhross 0.025
redmarketing 0.025
daveiam 0.026
lindydreyer 0.027
evesimon 0.030
johnmerritt 0.031
09ntc 0.034
smaclaughlin 0.036
apkat 0.045

Pagerank

pj_nccnp 0.008
nonprofit_tech 0.008
rachelannyes 0.009
andrewjcohen 0.009
johnmerritt 0.010
kalabird 0.011
09ntc 0.014
ntenhross 0.021
kanter 0.024
cshirky 0.048

Hubs

apkat 0.02
chadnorman 0.02
staceymonk 0.02
jaredpeters 0.02
darimonline 0.02
webb 0.04
npcommunicator 0.04
rachelannyes 0.06
kanter 0.06
smaclaughlin 0.13

Authorities

daveiam 0.01
nonprofit_tech 0.01
johnmerritt 0.01
09ntc 0.02
rachelannyes 0.02
webb 0.02
kanter 0.05
ntenhross 0.07
shaunsullivan 0.16
cshirky 0.23

Metadata

Total
3834
HT (heard through)
1
OH (overheard)
4
RT (retweet)
529

Update: Did a check on the closeness centrality within the largest strongly connected component. The leaders there are:

Closeness Centrality (in largest SCC)

lindydreyer 0.363
npnicole 0.364
andrewjcohen 0.365
ntenhross 0.368
09ntc 0.376
evesimon 0.379
kalabird 0.382
johnmerritt 0.395
smaclaughlin 0.418
apkat 0.420

I think this is related to apkat, smaclaughlin, johnmerritt, etc. being such good reporters across a number of people. They also have high outdegree centrality, which would indicate that. Anyway, these people are closer to all other people in the graph by dint of have the shortest average distance to all other nodes.

Copyright Mike Edwards 2006-2009. All content available under the Creative Commons Attribution ShareAlike license, unless otherwise noted.