←back to thread

142 points mzs | 1 comments | | HN request time: 0.258s | source
1. anigbrowl ◴[] No.19401727[source]
Is anyone aware of tools/research on instrumentation of such activity on social media? I don't mean text mining or semantic analysis of trolly content but the patterns of activity and interaction. At present I use directed graphs with k-means type clustering/community detection algorithms but that can lead to difficulty seeing the forest for the trees.

Imagine, for example, that Hacker News members were grouped in different teams - red, greeen, yellow, and blue - and that the teams competed to dominate discussion threads and grow their own discussions as semantic 'territory'. Imagine further that you had to represent this without any text, using only the time, post size, and threading structure, but with prior knowledge or manual curation of team membership. I'm particularly interested in how a thread could be represented spatially, but not as a time series (ie viewing progress over time would require animation). It's easy to visualize the ebb and flow of territorial conflict on a geographic map, with armies crossing borders and territories expanding or shrinking over time, but how to do that for emergent 'territory' like the expanding and branching threads of a busy discussion topic? I sometimes imagine it like a growing plant or tree with different colored leaves to represent competing affiliations, but there are surely better approaches.

tl;dr how can you visualize evolving discussion threads without using text?