Transcript Microsoft PowerPoint

NCRM 2012: DSR Showcase
Cardiff Online Social Media ObServatory
(COSMOS)
Matthew Williams, Pete Burnap, William Housley, Adam Edwards, Jeffrey
Morgan, Malcolm Williams, Omer Rana & Nic Avis
SOCSI/COMSC Research Network
School of Social Sciences & School of Computer Science and Informatics
Cardiff University
[email protected], [email protected]
Tweet comments #cosmoscardiff and follow @cosmos_cardiff
Context
• ‘Coming crisis of empirical sociology’ (Savage and
Burrows, 2007)
• Social media and the interactive web (Web2.0) provide
potential for systematic data mining and analysis of
naturally occurring data
• Development of a ‘social science digital tool kit’
Methodological Aim and Objectives
Aim:
• To evaluate the technical, methodological, and ethical challenges
presented by social media data in the context of social sciences.
Objectives:
• To demonstrate how APIs can be utilised within the COSMOS platform
to crawl, harvest, index and visualise qualitative and quantitative
social media data;
• To interrogate social media data using social science criteria for
quality and robustness;
• To examine the ethical and legal issues with harvesting data from
social media sources.
Demonstrator Aim and Objectives
Aim:
• To analyse social media data for the purposes of monitoring tensions
before, during and after major events (e.g. urban riots; industrial
action; political protests/elections; major sporting events etc.).
Demonstrator Objectives:
• To examine measures of connectivity in the analysis of social media
data in relation to tension indicators;
• To examine tension sentiment in social media data and their
correlation with events;
• To examine the feasibility of ‘mashing’ official curated data sources
with social media data.
Cardiff Online Social Media ObServatory
Consider process
sharing – similar to
Forms of analysis:
(1) Connectivity/Networks
(2) Frequency
(3)Sentiment
(4)Tension
(5)Anomalies
Social Media Tension Analysis Toolkit
• Started with sentiment analysis
– SentiStrength and Alchemy sentiment ‘tachographs’
• Developed an event-oriented tension scale (4 level ordinal
scale)
• Will lead to tension ‘tachographs’ and mixed method
(sentiment and tension)
• Study similarity/difference between S and T (can tension be
positive?)
• Tension colour coded word clouds in real/useful-time
• Network connectivity graphs
– Tension scale colour codes for nodes (users) and edges
(communications)
– Network capital/influence
• Anonymised qualitative tweet content data viewable
Tottenham Court Road Incident
Euro 2012
Mashing Crime Data with twitter
Next steps
• Evaluate Tension Ordinal Scale
– Police coded data
– Rule engine based on ordinal tension scale
– Machine Learning methods and Neural Networks
• JISC funding
– Develop platform with Manchester, St Andrews, Leeds, Leicester,
Wolverhampton and UCL
– Joint comparative case study on social media and crime, disorder and the
election of Police Crime Commissioners (Manchester and Cardiff)
– Requirements gathering
– Training and dissemination
Tweet comments #cosmoscardiff and follow @cosmos_cardiff