Statistical Analysis of the
Social Network and
Discussion Threads in
Slashdot: A HORRIBLE PAPER
presentation by Mykell Miller
Why study Slashdot?
● Because you haven't heard of:
– Non-nerds
– Youtube's comment ranking system
– Other news sites that allow comments
– Websites were you can't post anonymously
● Because of the important relationships on it
– People visit Slashdot to read articles, not comments
– Most comments are not discussions
The Social Network: Building the Network
● Over 30% of comments were discarded
– They're not studying the social network on
Slashdot. They're studying their favorite part of it.
– The discarded comments are not at all random –
discarding them greatly changes the structure of the
network
The Social Network: General Description
● The table on page 647 is horrible
– What kind of random graph? Erdos-Renyi? With
what probability of a link being formed?
– So many abbreviations that it's impossible to read
The Social Network: Degree Distributions
● Power law vs lognormal distributions
The Social Network: Mixing by Score
● Discarding -1 comments
– Don't you want to know how much spam, trolling,
etc. is on Slashdot?
– Or do you just not want to tarnish the image of your
beloved Slashdot?
● Bias in favor of good writers
– Duh
The Social Network: Community Structure
● Most clusters are of size 1
– Most people ignore each other.
● This section is determining the strength of ties
– But you ignored the research of Granovetter, the
daddy of strength of ties
Structure of the Discussions:
Radial tree representation
● Most branches are only 1 deep
– People usually don't reply to each other
– It's not much of a network, it's more of an
assortment
Structure of the Discussions:
The H-index as a Structural Measure of Controversy
● The worst section of a paper to ever be written
● Objective analysis of a subjective quality
requires the objective measure to make sense
● Or the objective measure can be based off of
something subjective but quantitative, such as
someone ranking the level of controversy on a
scale.
Structure of the Discussions:
The H-index as a Structural Measure of Controversy
● Good measures of controversy:
– The frequency of swear words in comments
– Sentiment analysis of comments
– The results of a survey in which people ranked the
level of controversy
● Barely passable measures of controversy:
– One person ranks the level of controversy
● Horrible measure of controversy
– The depth of the tree