Abstract for a talk I’m giving tomorrow some time between 1300h and 1500h at the University of Ottawa. Let me know if you want to attend (slides will be up later).
Follower / Following networks are essentially meaningless on Twitter due to the prevalence of spam. However by creating the graphs of conversation networks it is possible to create a better picture of more meaningful connections – the other users that interact / are interacted with by a given user. For power users, however, these graphs can be extremely busy, making it difficult to pick out the most important conversations and connections.
One potential way to summarize the most important connections in a network is to pull out cliques – completely connected sub-graphs. These cliques may represent part of a users core network, or a suggestion of new users to interact with by generating those cliques that a user is connected to. For example, user A might be in a clique with users B, C, D and users B, C, D may be in a clique with a 5th user, E. This suggests that user A might well be interested to interact with user E, as well. This may also help us determine tie strength as well, as a clique is likely an indication of a stronger tie strength than just a singly connected node.
One reply on “Applied Clique Finding: Discovering More About Your Twitter Network”
[…] Abstract can be found here. […]