Macroscopic Exploration of the Twitter Social Graph
Maksym Gabielkov, Arnaud LegoutEPI DIANA, Sophia Antipolis{maksym.gabielkov, arnaud.legout}@inria.fr
How information propagate?
What are the influence mechanisms?
Friends
ProducerConsumers
Follow Relationship in Twitter
Alice Bob
Bob follows AliceAlice follows Bob
The Twitter Social Graph
Alice Bob
+500 millions nodes+24 milliards edges
Challenges1. Collect the graph2. Decompose the graph3. Give a physical meaning
to the decomposition
8
How is constraint information propagation?
Identify the highways
3 34
1 1
4 1
11
1 1
1 1
1
3 34
1 1
4 1
11
11
1 1
1
250 millions
250 millions
7
26 000
Macrostructure of the Twitter social graph
SCC decomposition
Directed acyclic graph
LSCOUT
IN
DISCONNECTED
IN-TENDRILSOUT-TENDRILSOTHER
BRIDGES
Macrostructure of the Twitter social graph
Directed acyclic graph Macrostructure
Which physical meaning for the decomposition?
18
19
20
21
22
1% accounts
<0.01% tweets<0.01% edges
98% of the tweets98% of the edges50% of the accounts
Regular activity
1,5% of the tweets5,3% of the accounts0% outgoing edges
Selfi
sh ce
lebriti
es
21,4% of the accounts0,25% of the tweets
Passive users
21,6% of the accounts99% no edge80% no tweet
Spammeurs
Macroscopic Exploration of the Twitter Social Graph
Maksym Gabielkov, Arnaud LegoutEPI DIANA, Sophia Antipolis{maksym.gabielkov, arnaud.legout}@inria.fr
Following et follower
Alice Bob
following follower
Alice
following follower
Bob
Twitter in 2009 41.7 million users 1.47 billion follow links Average degree: 35 Partial crawls
Twitter in 2012 537 million users 23.95 billion follow links Average degree: 44 Complete crawl