July 10, 2008
Visualising CoAuthors in Open Repository Online Papers, Part 2
Having tinkered using Moowheel to visualise coauthors on references pulled from the OU's open repository (ORO) in Visualising CoAuthors in Open Repostory Online Papers, Part 1, to give a graphic which isn't really that useful or interesting (too cluttered), I thought I'd try out a couple of network visualisations using Many Eyes and Graphviz.
It's possible to construct graphs using both these tools by organising the data in pairs - so for each reference in a collection of references (for example, all the papers produced by a particular unit, or all the references returned from a title/abstract search on a particular query term), I pulled out every possible pairwise combination of authors. (So for exampple, a paper authored by A, B, C and D generates (A, B), (A, C), (A, D), (B, C), (B, D), (C, D) as all the possible pairwise coathor relations for that reference.) For convenience (i.e. as a result of laziness), I allowed duplicate pairs into the list (so if A and B had coauthored mutliple papers, their pairwise cauthor relation would appear many times in the data set.
So what's the result? Well here's a Many Eyes view over references produced from the Academic Unit/Department: Educational Policy, Leadership and Lifelong Learning:
It's possible to zoom in to the display and highlight or search for individual authors, and see who they have co-authored with:
As it seems that Many Eyes now allows visualisations to be embedded, let's see what happens if I try to embed the above.
At a high level, this visualisation allows us to pick out clusters of what we might refer to as different co-authoring teams within the unit. If we were visualising references returned from a search on a particular topic, the result might allow us to detect different groups working in a similar who don't author with each other, or different domains that use the same term but in completely different senses.
There are also a few things to notice in the detail about the Many Eyes network visualisation. Firstly, ambiguity in naming has a visual effect: WoodsPA and WoodsP are the same person I think, so their different representation creates an artificial degree of separation in the visualisation. Secondly, the node size reflects the degree of (I think) each author node - that is, the number of coauthors are connected to. I'm not sure if it also represents the absolute number of times the name appeared in the uploaded data set? Which leads neatly to the third point: the repeated occurrence of the same coauthor pairs are not apparently visualised.
If we now look at a fragment of a Graphviz directed graph (the direction goes from authors who appear earlier in the reference list to authors who appear later) we see that repeated co-author pairings are displayed - the number of connections between two authors represents the number of references they have coauthored:
Visualising the results of a search (e.g. a search over ORO for the term "mobile") reveals far more structure - here's a significant fragment of a Graphviz visualisation of couthors on references returned from just such a search:
Here's a more detailed view of part of the above:
So what? So nothing really... it just seemed like a good idea at the time, another way of trying to identify professional networks and teams based on their outputs... and it was relatively trivial (and quick) to do... ;-)
Tags: viz, oro, visualisation, graphviz, many eyes
Posted by ajh59 at July 10, 2008 11:26 AM