Odysseus

Twitter is one of the surprise success stories of the last couple of
years and today 10s of millions of users are happily sharing their
thoughts, opinions, and insights with their friends and followers, all
in the form of 140-charater short text messages.

Twitter's simplicity introduces some interesting challenges. For new
users it can be confusing about who to follow to get the most from the
service. And when you are up and running it can be a real challenge to
deal with the overload of messages that stream from your friends.
Wouldn't it be useful to develop a tool to allow you to better filter
your Twitter stream, so that you to notice interesting messages more
easily, saving you time and energy, and keeping you fully informed?

In this project the student will develop just such a tool by adapting
some existing technology developed in CLARITY to Twitter. In
particular we will look at how to evaluate the reputation of a Twitter
user by monitoring how other users respond to their tweets. For
example, if a user's tweets are often re-tweeted then this is a sign
that they are tweeting about interesting things. More re-tweets means
greater reputation and reputation can be used as a mechanism to weight
the tweets that come from different users.

The student will benefit from access to a number of CLARITY
researchers who are working on related issues. CLARITY already has a
comprehensive dataset of Twitter data from live users (millions of
tweets form hundreds of thousands of users) for example, and the
student enjoy the support of this team during the internship.

Demonstrable Outcomes/Workplan:
Weeks 1-2: Background Research. (brief description)

Weeks 3-4: Adapting the Reputation Model. The CLARITY team has
developed a reputation model for evaluating user reputation in a
social search setting (a la HeyStaks). This generic model can be
adapted to other settings where there is evidence of social
collaboration between users, such as in Twitter, where actions like
re-tweeting can be viewed as a form of collaboration. During this
phase the student will work to adapt the current reputation model for
use on Twitter data.

Weeks 5-6: Evaluating the Reputation Model. During this phase the
student will carry out a detailed evaluation of their reputation model
on a large-scale Twitter dataset.

Weeks 7-10: Development of reputation based ranking application. Here
the student will develop a simple Twitter client which incorporates
reputation based ranking (of tweets and users).

Supporting Material:
Twitter API (http://apiwiki.twitter.com/)

Kevin McNally, Michael P. O’Mahony, Barry Smyth, Maurice Coyle and
Peter Briggs (2010) "Towards a Reputation-based Model of Social Web
Search" In: 2010 International Conference on Intelligent User
Interfaces (IUI 2010), Hong Kong, China, 7-10 February 2010