Integrated Visual Analysis of Global Terrorism

1
Integrated Visual Analysis of Global Terrorism

• Remco Chang
• Charlotte Visualization Center
• UNC Charlotte

2
Integrated Terrorism Analysis
Multimedia
Real Time
Known Events
Visual GTD
3
Video Analysis Goals

• to describe trends in news content over time
• to discover breaking news and hot topics over
  time
• to trace conceptual development of news
• to retrieve news of interests effectively
• to collect evidences and test hypotheses for
  intelligent analysis
• to compare group (such as different channels)
  differences in content
• to associate news content with social events

4
Multimedia Analysis
5
Video Analysis Example

• CNN Fox News
  MSNBC

• News contains view points and opinions
• Find local, regional, national, and international
  reports of the same event to get a complete
  picture

6
NVAC Collaborations

• PNNL A. Sanfilippo (Content Analysis and
  Information Extraction of closed caption)
• PNNL W. Pike (Emotional state extraction from
  closed caption)
• Penn State A. MacEachren (Geographical
  analysis)
• Georgia Tech J. Stasko (Jigsaw, entity
  relationships)
• Visual Analytics is the point of integration!!

7
Integrating Terrorism Data Analysisand News
Analysis
Terrorism Visual Analysis
Terrorism Databases
Terrorism VA
Jigsaw
NVAC
Stab/ TIBOR Reasoning Environment
Framing, Affective Analysis
Broadcast VA
News Visual Analysis
News Story Databases
Next full, Web-based multimedia content
8
Visual GTD Flow Chart
Entity Relationships (Geo-temporal Vis)
Dimensional Relationships (ParallelSets)
Entity Analysis (Search By Example)
9
WHO Terrorist Groups
Five Flexible Entry Components
What
WHERE
WHEN
10
Enter System by single or multiple Selections
System will supply Specific Information
Drilldown to Original Info
11
Parallel Sets View

• Parallel Sets
• Displays relationships among categorical
  dimensions
• Shows intersections and distributions of
  categories

12
Parallel Sets View

• Dynamic filtering on continuous dimensions can
  show more information
• Here we see the large proportion of facility
  attacks and bombings in Latin America during the
  early 1980s

13
Analysis using Longest Common Sequence (LCS)

• Two strings of data (each representing a series
  of events)
• GATCCAGT
• GTACACTGAG
• Basic algorithm returns length of longest common
  subsequence 6
• Can return trace of subsequence if desired
• GTCCAG
• GATCCAGT
• GTACACTGAG
• Additional variations can take into account event
  gap penalties, time gap penalties, and
  exploration of shorter, or alternate, common
  subsequences

14
Grouping using MDS in 2D

• Each o represents a terrorist group
• Groups form cluster according to naturally
  occurring trend sizes
• Sharp divide between large clusters in right
  hemisphere
• Left hemisphere contains many smaller clusters

MDS Analysis by TargetType