Twitter 1 Ramine Tinati, 1 Leslie Carr, 1 Wendy Hall, 2 Jonny - - PowerPoint PPT Presentation

Identifying Communicator Roles In Twitter

1Ramine Tinati, 1Leslie Carr, 1Wendy Hall, 2Jonny Bentwood

24th March 2012

1Department of Web and Internet Science 2Edelman Ltd

Presentation Outline

2

• Background Research and Inspiration

– Existing and relevant literature

• An Overview of Twitter

– The Influence Layer

• Topology of Influence

– Defining a new model to identify different communicator roles in Twitter

• Implementation and Findings

– A review of the implementation – Evaluation of the Topology Of Influence

Background and Inspiration

3

• A growing body of literature investigating the Twittersphere

– Event detection – Crisis management – Identifying different user characteristics within networks – Spam detection and prevention – Propagation and Cascading of Information

• Inspiration: It is changing the way society is communicating and
• perating

– As part of the long term Web And Internet Science Project - The Web Observatory – Joint project with Personal Relations company, Edelman

Research Questions…

4

• Lots of interest recently in the flow of messages within the twittersphere

– Hashtags such as #UCDavis, #Occupy, #Obama #DrWho, etc. quickly become trending topics with a large number of users tweeting

• Is all the information on Twitter valuable? Obviously not…

– 60% of Tweets are considered Spam (Pear, 2009)

• But do these large network of users and tweets truly reflect the

conversations occurring? – Are all users equally contributing to the conversation, or is it just a select few?

Can we identify different communicator roles in a Twitter conversation?

Twitter - Overview

5

• Twitter: a micro-blogging service

– Users can post 140 tweets to each other – These tweets can contain:

• Hashtags
• Other users (@joeblogs)
• URI/URLs
• Twitter also allows for sharing of

messages – Via the Retweet function – A social practice of sharing ideas

Social Network Messaging System (centralised) Web Database (Cloud database)

Implementation

{Undifferentiated users exchanging messages} {Firehose view of the messaging, API connections} {Storage of the messages, invisible for public access}

Twitter Implementation Stack Twitter Retweet Functionality

U1 U2 U3 U4 U5 retweet U4 Retweet U4

Communication on Twitter

• It provides a medium for people to communicate

– Information can be spread extremely fast

• But so can rumours and spam
• It is facilitating change

– Businesses, Organisations, Governments, Individuals

• Just like in the real-world, information tends to be repeated if it is

important or valuable (Boyd et al., 2010) (Cha et al., 2010) – This is the same in Twitter, tweets become retweeted

Twitter – Adding an Influence Layer

7 Social Network Messaging System (centralised) Web Database (Cloud database)

Implementation

{Undifferentiated users exchanging messages} {Firehose view of the messaging, API connections}

Influence Network

{Different user roles in socially mediated communications} {Storage of the messages, invisible for public access}

Twitter Implementation Stack + Influence Layer

Topology of Influence (TOI)

8

• A User Classification Model

– Based on Edelman’s Professional Experiences in the PR Industry

• Five Different Categories of Users

– Each have their own behavior and communication characteristics

• A pyramid of user roles

– Similar to Everett Roger’s Diffusion of Innovations (1962)

TOI Example - Idea Starter vs. Amplifiers

Name Influence NHS 29.51 WHO 21.51 Department of Health 18.22 The European Union 15.07 FSA 12.44 British Medical Journal 10.65 The Scotish Executive 10.05 BBC 9.77 NICE 9.42 NIH 7.63 The National Obesity Forum 7.30 BHF 7.21 DFES 6.62 Health Protection Agency 6.23 NAO 6.02 OFCOM 5.94 British Medical Association 5.53 Audit Commission 5.42 OFSTED 5.14 Sport England 4.80 Jamie Oliver 4.74 British Nutrition Foundation 4.62 Guardian 4.13 Dept for CMS 4.02 .00 .91 .82 .78 .75 .71

9

Implementation of TOI

10

• Implement the Influence Layer using the TOI

– Like other research, use the retweet network to determine influence

• Using the TOI to identify different communicator roles in a Twitter

retweet network – Idea Starter – Users with a large number of retweets – Amplifiers – Users who are first to retweet an Idea Starter – Curator – Users who retweet more than one Idea Starter – Commentator – Users who occasionally retweet an Idea Starter

ReFluence – Combining Twitter and TOI

11

• A tool to visualize and analyze the different TOI communicator

roles – Uses harvested Twitter data to provide a temporal view of the network of retweet messages

• Calculates a number of basic Social Network Analysis Metrics

– Network Statistics – Nodes, Edges, Degree

• Provides a visual exploration into the communications between

users – Visually tracing the dynamic changes in the network structure

ReFluence and TOI - Example

12

Evaluation

13

• Does the TOI provide anything new?

– Validate the different communicator roles

• Evaluation was performed by using different Twitter datasets ranging

in – Topic – Size – Geographic context

Dataset Tweets Retweets Users #twilight 529530 139441 336446 #DrWho 709093 204301 104688 #SOPA 1004482 438894 485692 #Occupy 41568 16673 29025 #OccupyLondon 19128 9834 7548 #Nov9 12831 7188 4737 #Nov30 22054 14243 12330 #UCDavis 7950 3895 4523

Findings

14

• Power law underpins the distribution of retweets amongst users

– The distribution of users were similar across all different datasets, irrespective of topic

• Similarities between other user roles (curators and amplifiers) across

datasets

0.0001 0.001 0.01 0.1 1 10 100 1 10 100 1000 10000

Distribution of Users (log) Number of Retweets (log) DrWho Occupy N30 OccupyLondon Nov-09 UCDavis twilight SOPA

Future Work

15

• Further Investigation and validation of the TOI
• Message propagation as a possible path to determining influence
• Examine the scale-free / power law properties of the retweet network

Thank you. Questions?