Summary by Patrick

The goal of this paper is to identify people with influence on the social
network Twitter.

A Twitter user with influence is defined as a user who posts a URL which
diffuse  through the follower graph. The study is limited to those users who
"seed" content, which mean they post a URL that they have not seen through the
follower graph.

To calculate the influence of a URL the authors tracked the diffusion of the
post from its seed node throughout the follower graph until either the cascade
terminated, or the diffusion.

Then in order to evaluate the URL and to identify the content, they first
remove those posts they knew to be spam or in a foreign language and grouped
the remaining URLs in a logarithmic bin and select randomly from this. The
randomly selected URLs were analysed by human subjects who classified the
URLs.


Since this paper was written twitter has added a retweet feature which allows
you to retweet at the push of a button. This should allow a new study of this
topic to be done with additonal data. Being able to track retweets from a
single seed would add additional data.

Summary by Zahid

Everyone¡¯s an Influencer: Quantifying Influence on Twitter
Eytan Bakshy, Jake M. Hofman, Winter A. Mason, and Duncan J. Watts

Summary


In this paper authors try to investigate the attributes and relative influence
of 1.6M Twitter users by tracking 74 million diffusion events. During the
study it found that the largest cascades tend to be generated by those users
who have been already influence in the past, those users have a large number
of followers so this is the main reason of influence. But it also find that
those predictions related to URL or some particular users generate large
cascades are not valid or unreliable.


Twitter ecosystem is very good system because it is well suited to studying
the role of influencers. The ordinary individuals play a vital role to
influence on others, like experts, journalists, semi-public figures, actors,
role models, highly visible public figures and government officials. These
individuals have capability to influence the number of people through
different channels like celebrity endorsing a product on television or in a
magazine advertisement. So these individuals exerts more influence on people
than a trusted friend endorsing the same product in person. Authors also noted
that Twitter is very special case because in a large cascades the observations
are rare likely to apply in other context as well. There is one more term
called ¡°word-of-mouth¡± is spread via many small cascades, like marketers,
planners, and celebrity. So they get the better data about potential
influencers over extended intervals of time.

----- BEGIN SUMMARY ------

Measurement and analysis of online social networks
summary by guangyu han
------------------------------


The paper uses a crawler to crawl through online social networks such as
Youtube, Orkut, Flickr and LifeJournal. The author collect a data set by
crawling cover the following number of users: 1.8 million out of 6.8 million
(26.9%) for Flickr, 5.2 million of 5.5 million (95.4%) for LiveJournal, and
3.0 million out of 27 million (11.3%) for Orkut. it gives us a tangible view
of properties in large social networks

Many of the properties in the paper were already proposed before Adamic et Al
[3]  found small-world behaivor, and local clustering. Kumar et Al. [26] found
a large strongly connected component.

All networks follow a power-law is propotional to K^(-i). Small-world
property. Very short path length Scale-free means  nodes with high degrees are
connected to other nodes with high degrees

questions what is the  influence of fake users  when performing the  analysis?
and How to avoid  these influence .

------END SUMMARY-----------