On Saturday morning I woke up and as usual checked my twitter feed to discover a tweet from renowned physicist Lisa Randall saying she will talk at ESOF. I had no idea what ESOF stands for, so I started doing some research on the subject, leaving me even more positively surprised.
ESOF stands for Euroscience Open Forum held biennially in different Europen cities, dedicated to scientific research and innovation.The Forum is a meeting place for scientists, researchers, business, innovators, policy makers, technology experts and everybody from all over Europe who is interested to learn more about scientific discoveries and research in fields such as mathematics, music, geography, genetics, physics, etc. ESOF 2012 was being held in Dublin, Ireland and one of the hot topics was the recent discovery by CERN which is believed to be Higgs boson particle.
I furthermore found out that ESOF is live streaming some of the lectures (one of them being of prof. Randall’s later that Saturday). So I’ve started watching the conference and looked through the live tweeting on the right side of the streaming. This made me wonder how the tweeting community of ESOF looks like and what I might see if I do a social network analysis on the #esof2012 community. Additionally, I just wanted to brush up my NodeXL skills.
I used NodeXL, during 15.07.2012, setting the search on twitter search network for the hashtag #esof2012 and limiting the data to 500 people. I decided to limit the people mostly because my NodeXL sometimes shows error if I do a search with over 500 people. Especially often this happens since Twitter put some restrictions on data gathering.
In the graph above you can see not only who follows whom, but also the conversations between the twitter users mentioning #esof2012 hashtag. Connection are created when users follow, mention, reply one another. You can also see there are 500 vertices (twitter users) with 2907 unique edges and 1178 duplicated edges.
The graph layout is using Fruchterman-Reingold algorithm and the type is: directed. I have also grouped the users by color through Wakita-Tsurumi clustering algorithm. Additionally, betweenness and closeness centralities and eigenvector centrality were calculated. Betweenness centrality basically shows us the measure of the vertex’s ability to be a bridge between other people in the network. After all the conference covers wide variety of topics and interests, so I’ve wanted to check if the different scientific groups are isolated, or there are people bridging the clusters.
Closeness centrality measures the average distance between a vertex and every other vertex in the network. The lower closeness centrality is better because it shows that a person is directly connected to most other in the network.
Through Eigenvector centrality I wanted to measure the influence of a node in a ESOF2012 twitter search network and establish the most influential twitter accounts.
The Average Betweenness Centrality is 562,900 while the Average Closeness Centrality is 0,011. The Average Eigenvector Centrality is 0,002.
Here are as well some overall graph metics:
|Overall graph metics|
|Edges With Duplicates||1178|
I furthermore wanted to see the groups of users and for that I used the layout style function to lay each of the graph’s groups in its own box with hidden intergroup edges with strength of the repulsive force between vertices set to 3.0 and iterations per layout 10.
I also played with the functions to see how the groups are connected:
As final words, I am very happy I found out about ESOF – I watched Lisa Randall’s presentation, then I saw the Debate on Scientific Publishing and Open Access and the Keynote address from Rold-Dieter Heuer on “The search for a deeper understanding of our universe at the LHC: The World’s largest particle accelerator”.