One of the obvious problems with analysing only one conversation in the paper is that it’s difficult to say how representative is that one for the community under study, so I wanted to use the advantage of having tools to do more automatic analysis now to say if there were other similar conversations in the community.
Basic facts of the initial conversation:
- mapped manually, by following outgoing links and trackbacks
- includes self-linking posts (those that are not linked to/from other bloggers)
- two languages (English and German)
- 23 Nov 03 – 18 Jan 2004
- 17 blogs, 30 posts, 59 comments (only 1 post and 3 comments in 2004)
- 32 people (27 of whom are bloggers; other 5 – not clear)
So what do we have now:
- weblog conversations mapped automatically, by extracting groups of posts that link to each other from
- 34 weblogs in English (see Weblog conversations revisited: an introduction on dataset and caveats)
- 1 Jan – 31 Dec 2004
While experimenting with mapping the conversations we tried two approaches: (1) only focusing on links between posts of different weblogs and (2) including self-linked posts. Obviousely, going with the second choice is needed for compatibility with the original conversation. However, when trying to extract all groups of linked posts we run into a problem: apart from several small conversations, we would get one of 1028 posts with all 34 blogs participating (I’ll blog on possible explanations later).
- extract groups of posts that do not inlude self-linking
- add self-linking posts only to those in a conversation
So, those would look more or less like those pictures on the right, where pink blockes are self-linked posts added (scale and colors have changed between iterations, this is just for an idea).
An overview of what we’ve got is below (size of the bubble represents number of conversations with X blogs and Y weblog posts). From total of 182 conversations, most are very small (2 blogs, 2 linked posts), but there is a number of bigger ones. Our originally studied conversation would be in the top right area of the graph (~ conversations of that scale do happen, but not that frequently).
If you look at the graph, you can see that most of the conversations on the right shift up (number of posts in a conversation increases due to self-linking). I guess it indicates that the more complex a conversation becomes (more weblogs, more posts), the more likely it would be connected to other (=not belonging to the conversation) posts of bloggers who participate. So, it seems that people tend to get inspired by complex conversations in their future thinking (or at least find it useful to link to later on).
Archived version of this entry is available at http://blog.mathemagenic.com/2007/08/14.html#a1934; comments are here.