DEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
Vra 2013 cultural heritage data visualizations sula
1. Visualizing cultural networks
Chris Alen Sula
School of Information & Library Science, Pratt Institute
4 Apr 2013 – Visual Resources Association
Session #11: Cultural Heritage Data Visualizations
2. Overview
‣ Case study: Occupy Wall Street Project List
‣ What networks can represent
‣ Data structure for networks
‣ Gephi network software
‣ A few recipes
Cultural heritage documents often contain
information about relationships—the types of
relationships that can be explored and studied
through network visualization.
3. Case study: Occupy Wall Street Project List
Project information
Structured data
DAP <—> Occupy Town Square
DAP <—> OWS Direct Action
DAP <—> OWS Silkscreen Guild
. . .
Occupy Town Square <—> OWS Direct Action
Occupy Town Square <—> OWS Silkscreen Guild
. . .
4. ―less than a year after the last protester was
removed from New York City's Zuccotti Park, the
movement has re-emerged as a series of laser-
focused advocacy groups that, loosely
organized under the Occupy umbrella, are trying
to effect change in a variety of sectors, financial
and otherwise.‖
Time Magazine, Dec 3, 2012
5. Case study: Occupy Wall Street Project List
Feb 2012 Apr/May 2012 June/July 2012
6. What networks can represent
‣ kinship and personal connections (friends, partners, co-
performers, colleagues, acquaintances)
‣ organizations (roles, partnerships, alliances)
‣ linguistic associations (words, topics)
‣ trade routes, voyages, infrastructure
‣ communication (letters, social media)
‣ ideological ties (claims, theories)
‣ In each case, we look for similar things (nodes)
which are related in regular ways (edges)
7. Data structure for networks
‣ edge table*
‣source* / target* / direction / weight / timespan
‣ node table
‣id* / name* / property 1 / property 2 / etc.
*required
8. Data structure for networks
‣ Look for consistently recorded information (recurring
entities, similar connections)
‣ Think hard about whether connections are directed or
undirected—this may change the structure
‣ Expect that you‘ll have to resolve some problems with
messy data (ambiguity, variant spellings, etc.)
‣ Search for familiar real-world objects to make into
nodes (e.g., people) and treat relationships between
them as edges (e.g., being displayed together at the
same gallery event)
9. Nodes v. Edges
‣ Networks can be single mode (one type of node, e.g.,
people) or multi-mode (e.g., people and institutions)
‣ Most layouts and network statistics are built for single
mode networks, including ones in Gephi
‣ In many cases, it is advantageous to
‣ attribute-ize node properties (e.g., make ‗person‘ and
‗institution‘ values in a ‗type‘ attribute)
‣ edge-ize abstract ―things‖ that actually only connect
nodes (e.g., being included in the same catalog)
‣ rather than treating everything as a new type of node.
10. Gephi network software
‣ free and open-source for
Windows, Mac, and Linux
‣ allows for detailed design
adjustments (sizing, coloring,
filtering, labeling)
‣ community-developed plugins
provide additional layout
options and customization
‣ computes network statistics,
detects clusters
‣ exports files as image (PNG,
PDF), vector (SVG), or web
formats
http://gephi.org
11. A few recipes
‣ Categorization
Map terms that appear throughout a subject vocabulary by treating
hierarchy as an edge and each term as a node. (Reveals conceptual
structure of categorization system.)
‣ Collaboration
Map people that collaborate on works by treating each person as a node
and each instance of collaboration (i.e., work) as an edge that connects
them. (Reveals social patterns.)
‣ Provenance
Map provenance relations between works by treating each work as a
node and co-location/co-exhibition as a relationship that connects them.
(Reveals curitorial relationships.)
12. Contact
Chris Alen Sula
Assistant Professor
Pratt Institute, School of Information & Library Science
http://chrisalensula.org
@chrisalensula on Twitter
csula@pratt.edu