[PPT] - Jean-Philippe Cointet CREA (CNRS/EP , France) AND Camille Roth PowerPoint Presentation

SLIDE 1

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Socio-semantic dynamics in a blog network

Jean-Philippe Cointet

CREA (CNRS/EP , France)

AND

Camille Roth

CAMS (CNRS/EHESS, France)

IEEE SOCIALCOM09, VANCOUVER, BC – AUG 29–31, 2009

SLIDE 2

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

A SOCIAL network

Three kinds of links for each blog... citation: post citation links interaction: comment links affiliation: blogroll links ...where contents circulate in terms of topics (W) in terms of cultural items (U)

Dataset: US blogosphere

scope: 4 months of ’08 campaign network: citations nodes: 1, 066 blogs (RTGI)

citation link

A socio-SEMANTIC network

Three kinds of links for each blog... citation: post citation links interaction: comment links affiliation: blogroll links ...where contents circulate in terms of topics (W) in terms of cultural items (U)

Dataset: US blogosphere

scope: 4 months of ’08 campaign network: citations nodes: 1, 066 blogs (RTGI)

semantic characterization

“relevant” syntagms

(“health insurance”, “climate change”, “national security”, “super Tuesday”, “human rights”...)

urls: “www.youtube.com/x1hqwkeac”, etc.

SLIDE 4

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

A DYNAMIC socio-semantic network

Three kinds of links for each blog... citation: post citation links interaction: comment links affiliation: blogroll links ...where contents circulate in terms of topics (W) in terms of cultural items (U)

Dataset: US blogosphere

scope: 4 months of ’08 campaign network: citations nodes: 1, 066 blogs (RTGI) http://presidentialwatch08.com/

SLIDE 5

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Socio-semantic configuration

Jan 1 Michigan Primary Super Tuesday feb 17 0.01 0.02 0.03 0.04 0.05 0.06

time frequency of terms

super tuesday michigan california huckabee

d c b a 19/02 20/02 20/02 26/02

SLIDE 6

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Socio-semantic configuration

semantic profile of a blog i:

ˆ Wi (w) := Wi (w) P|W|

w=1 Wi (w)

· log |B| |{j, Wj (w) > 0}|

semantic distance between blogs i and j: δ(i, j) = 1 − ˆ Wi · ˆ Wj ˆ Wi ˆ Wj

[0;.1[ [.1;.2[ [.2;.3[ [.3;.4[ [.4;.5[ [.5;.6[ [.6;.7[ [.7;.8[ [.8;.9[ [.9;1] 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2

P (δ) δ

Semantic distance distributions. Triangles: computed over the whole set of possible blog pairs. Crosses: distribution computed on linked blogs.

SLIDE 7

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Computing link creation propensity

→ estimate the “propensity of interaction” ...that it is more or less likely for a node (or a dyad) with property “m” to receive a link ...which may be simply estimated by: ˆ f(m) = ν(m) N(m) ν(m) = number of links pointing towards an agent of type m

(resp. number of new dyads of type m) during a time period,

N(m) = number of agents (resp. of dyads) of type m.

SLIDE 8

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Computing link creation propensity

→ estimate the “propensity of interaction” ...that it is more or less likely for a node (or a dyad) with property “m” to receive a link ...which may be simply estimated by: ˆ f(m) = ν(m) N(m)

50 100 150 200 10 10

1

k ˆ f(k)

ν(m) = number of links pointing towards an agent of type m

(resp. number of new dyads of type m) during a time period,

N(m) = number of agents (resp. of dyads) of type m.

SLIDE 9

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Dynamics of the social network

in-degree effects

→ increasing, plateauing

topological distance effects

→ strong trend to repetition and local

interaction

semantic distance

→ strong trend to homophily

primarily “social”?

social distance and degree

50 100 150 200 10 10

1

k ˆ f(k)

SLIDE 10

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Dynamics of the social network

in-degree effects

→ increasing, plateauing

topological distance effects

→ strong trend to repetition and local

interaction

semantic distance

→ strong trend to homophily

primarily “social”?

social distance and degree

1 2 3 4 >4 10

−2

10

−1

10 10

1

10

2

d ˆ f(d)

SLIDE 11

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Dynamics of the social network

in-degree effects

→ increasing, plateauing

topological distance effects

→ strong trend to repetition and local

interaction

semantic distance

→ strong trend to homophily

primarily “social”?

social distance and degree

[0;.1] ].1;.2] ].2;.3] ].3;.4] ].4;.5] ].5;.6] ].6;.7] ].7;.8] ].8;.9] ].9;1] 10

−1

10 10

1

10

2

δ ˆ g(δ)

SLIDE 12

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Dynamics of the social network

in-degree effects

→ increasing, plateauing

topological distance effects

→ strong trend to repetition and local

interaction

semantic distance

→ strong trend to homophily

primarily “social”?

social distance and degree

1 2 3 >3 [0;0.2[ [0.2;0.4[ [0.4;0.6[ [0.6;0.8[ [0.8;1] 10

−4

10

−3

10

−2

10

−1

10 semantic distance δ social distance d propension p(d, δ)

SLIDE 13

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Dynamics of the social network

in-degree effects

→ increasing, plateauing

topological distance effects

→ strong trend to repetition and local

interaction

semantic distance

→ strong trend to homophily

primarily “social”?

social distance and degree

1 2 3 >3 50 100 10

−2

10

−1

10 10

1

10

2

social capital k social distance d propension p(d, k)

SLIDE 14

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Information flows: measures on the post network

Dyadic measures:

raw, weighted network, aggregated on 4 months attentional matrix a... → and total attention αa = 5/6 detachment matrix

“edge range”:

quantifying shortcuts

d b c a 1 3 2 2 1 3 f e 3 2 2 1 1 1 1

SLIDE 15

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Information flows: measures on the post network

Dyadic measures:

raw, weighted network, aggregated on 4 months attentional matrix a... → and total attention αa = 5/6 detachment matrix

“edge range”:

quantifying shortcuts

d b c a 1/6 3/4 2/3 2/6 1/5 3/6 f e 3/5 2/3 2/2 1/3 1/4 1/3 1/5

SLIDE 16

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Information flows: measures on the post network

Dyadic measures:

raw, weighted network, aggregated on 4 months attentional matrix a... → and total attention αa = 5/6 detachment matrix

“edge range”:

quantifying shortcuts

d b c a 6 4/3 3/2 3 5 2 f e 5/3 3/2 1 3 4 3 5

SLIDE 17

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Information flows: measures on the post network

Dyadic measures:

raw, weighted network, aggregated on 4 months attentional matrix a... → and total attention αa = 5/6 detachment matrix

“edge range”:

quantifying shortcuts

d b c a 6 4/3 3/2 5 2 f e 5/3 3/2 1 3 4 3 5

SLIDE 18

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Information cascade

d c b a 19/02 20/02 20/02 26/02

An example of diffusion subgraph, a common “resource” and a set of citation links between blogs

Diffusion subgraphs

10 10

1

10

2

10 10

1

10

2

10

3

10

4

size of diffusion subgraphs number of diffusion subgraphs

links nodes

⇒ heterogeneous cascade sizes

SLIDE 19

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

An ego-centered perspective

! " # $ %&'() )('() )('() )*'()

We focus on the total number of “transmissions” generated by blogs with a given total attention α

a bit more “global”...

second transmissions: we focus on “later transmissions”, i.e. after a first transmission event

role of the total attention on the number of diffusion links

10

−2

10

−1

10 10

1

10

2

10

−1

10 10

1

10

2

Total Attention α Number of tranmissions

Larger active readership => larger number of diffusion links, yet not linearly

SLIDE 20

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

An ego-centered perspective

! " # $ %&'() )('() )('() )*'()

We focus on the total number of “transmissions” generated by blogs with a given total attention α

a bit more “global”...

! " # $ %&'() )('() )('() )*'()

second transmissions: we focus on “later transmissions”, i.e. after a first transmission event

role of the total attention on the number of diffusion links

10

−2

10

−1

10 10

1

10

2

10

−1

10 10

1

10

2

Total Attention α Number of tranmissions

Larger active readership => larger number of diffusion links, yet not linearly

SLIDE 21

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

A more global perspective

→ role of edge-range on the number of grand-children

We focus again on transmissions occurring after a first transmission event

d c b a 19/02 20/02 20/02 26/02

10 20 30 40 50 60 70 80 90 100 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

edge range r Mean number of 2nd transmissions

An information which has been transmitted through a “median” link generates a larger number of grandchildren

SLIDE 22

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Concluding remarks

Co-evolution of content and relationships Patterns not necessarily linked to authority only Patterns not necessarily ego-centered only → divergent from the “neighbor-based-influence” perspective

SLIDE 23

The “blogosphere” as a socio-semantic network Link creation dynamics Diffusion dynamics

Concluding remarks

Co-evolution of content and relationships Patterns not necessarily linked to authority only Patterns not necessarily ego-centered only → divergent from the “neighbor-based-influence” perspective Thank you!