1. AGORA
Creating the Historic Fabric for
Providing Web-enabled Access to
Objects in Dynamic Historical
Sequences
ISAB 2012 Site Visit
13-Dec-2012
2. Overview
• Team
• Goals and Key results
• Events Narratives
• Event Descriptions in Agora
• Demos
• Pilots
• Dissemination Collaborations
• Reflection
3. Science Team
• VU History department
• Susan Legêne
• Chiel van der Akker
• Computer Science
• Guus Schreiber
• Lora Aroyo
• Marieke van Erp
• Lourens van der Meij
4. Heritage Team
• Rijksmuseum Amsterdam
• Geertje Jacobs
• Beeld en Geluid
• Johan Oomen
• Also part time PhD candidate in Agora
5. develop a social platform in which museum objects
are placed in an explicit (art)historic context in
order to provide a more complete illustrated
description of historical events, which finally is
complemented by the user-created narratives to
support reflections on the meaning of digitally
mediated public history in contemporary society
AGORA Goal
7. AGORA in numbers
• 15 papers, posters • 2 demos, 3 collections,
presentations
2 platforms
• 12 (inter)national • 2 BSc students
conferences
(graduated)
• 1 book chapter (in • 4 MSc students (3
review)
graduated)
• 5 workshops
• 1 history intern at the
• 2 panels
Rijksmuseum
• 2 Best Poster awards
8. Key Contributions
• Digital Hermeneutics (WebSci 11, best paper nomination)
• combining cultural heritage, Web the public
• analysis of historical narratives
• reflections on the evolving notion of historical events how user
topics become essential for their understanding
• Automatic Heritage Metadata Enrichment with Historic
Events (MW2011, paper)
• stream of use cases, methods techniques
• events community, panels workshops
• The notion of events historic events (Museum
Transfiguration (book under review)
• thesaurus of historic events
19. Conclusion: events are not
precisely circumscribed
• Both lay people and experts don t usually
agree on what events are
• Events carry different perspectives
• The vagueness of events is part of their
semantics
• The perspectives are also part of their
semantics
21. Antoine Watteau, Pélerinage à l'île de Cythère, 1717/18
Oil on canvas (129 x 194 cm), Schloss Charlottenburg, Berlin
22. his Pilgrimage to Cythera, based directly
on the imagery and ideology of two
ballets produced at the Opéra and
related works at the théâtre de la foire,
may be seen as the most complete
expression of an operatic, proto-
Enlightenment vision of an alternative,
utopian society .
G. Cowart, The Triumph of Pleasure. Louis XIV the
Politics of Spectacle, (University of Chicago Press,
2008), xv.
23. Description
vs.
Narration
----------------------------------------------------------------------
Identification
vs.
Significance
Re-description
vs.
Retroactive alignment
(stabbing as killing)
vs.
(stabbing as end of the Republic
rather than its renewal)
(painting as rococo-painting)
vs.
(painting as exemplifying proto-
Enlightenment)
Documentation
vs.
Exhibition
30. Pattern-based Event Extraction
• Seed list of 400 events
• Find text snippets on the Web around seed
events
• Consolidate snippets and extract patterns
• Result: 503 scored patterns (1132 pattern
variances):
• @ takes place and @ took place
(variations in style)
31. Pattern-scores: examples
PATTERN
E-‐nr
UC
A
B
TC
UC*TC
veldslagen|uit|de|
8
0.014
1438
2520
0.570
0.007
verwoest|Ejdens|de|
21
0.031
378
1600
0.236
0.007
Ejdens|de|
11
0.011
783923
35400000
0.022
0.0002
het|einde|van|de|
4
0.004
50770
6120000
0.008
0.00003
ontkennen|van|de|
3
0,001
3176
7480
0.424
0.003
was|voor|de
3
0.003
967
1590000
0.0006
0.000001
E-‐nr:
number
of
unique
events
in
the
first
1000
snippets
(predefined
list
of
400)
UC:
#unique
events/#snippets
(max
1000)
A:
total
occurrences
of
the
events
with
the
paYern
B:
frequency
of
the
paYern
TC:
A/B
high
score:
combinaEon
is
specific,
not
much
noise
expected
low
score:
paYern
combines
with
many
other
possible
non-‐events
UC*TC:
product
of
both
scores.
High:
relevant
and
highly
specific,
but
oen
not
frequent.
32. Results
• 2444 event candidates checked manually.
• Precision 56.3%
• Examples of erroneous event candidates:
• Mouths of Dutch girls
• German occupiers vs
German occupation
• Examples of incomplete events:
• Revolution of 12
• First battle of
33. Events: a computational
perspective
• Instantiated Event Types
• Sortal nouns with a PP and NE: Battle of Stalingrad, Death
of John Lennon
• Normalised verbs with a PP and NE: Excavation of Troy,
Election of Obama
• Referential adjective with an event type and a named
entity: American invasion of Iraq
• Named Events as Proper Names
• Transparent proper names: Great War
• Opaque proper names: 9/11, Pearl Harbour
34. Mismatch target text
• In 90% of museum records event name is not
mentioned
• but, often mention of participants, locations times
⇒ Create event instances from (semi-)structured
sources
⇒ Link to records through participants, locations,
times
41. Demo Functionalities
• Event browsing through themes, events structure object
and event properties; RMA and BG sub-collections;
alignment of their vocabularies; using the SEM model for
enrichment with event information
• Collection browsing event browsing and object clustering,
as well as through a map or a narrative view
• Proto-narratives are auto-generated based on user s
navigation path and organized in types, e.g., a topological,
conceptual, or biographical
59. Pilot 1:
University History Students
• VU History department
• 13 undergraduate history students
• volunteered to do demo assignment as part of
Cultural Sources of Political History course
• Students formulated a research question and tried
to answer it with the demo
• 2 groups:
• Group A used demo with narratives
• Group B used demo without narratives
60. Pilot 1:
University History Students
• Analysis of student reports Survey
• Reports specified the research question and answer found
• Students specified which pieces of information came from
the demo, or from other sources
• Students were able to correctly identify relevant
objects and historical context
• Proto-narratives helped form overarching narrative
• Majority would like extra information
• Demo did encourage different search approach
61. Pilot 2:
Secondary School Students
• Oelbert gymnasium Oosterhout
• Two classes (44 pupils) 3rd year secondary school
• As part of theme Indonesia in the history lesson they
had to answer one of two research questions
• What was the view of the Dutch Indies population on the
police actions ?
• What was the view of the Dutch population on the police
actions ?
• Assignment was done in class
• Results collected through click logs and
online survey
62. Pilot 2:
Secondary School Students
• Results
• 64% were able to correctly identify relevant objects to
their research question
• 41% were able to situate the objects in their historical
context, pairs answering the Dutch perspective question
did better on this
• proto-narratives were not perceived as very useful
• pupils expressed need for extra information in the
demo
63. Pilot 3:
Remembrance Community
• focus group 6-person (3m, 3f) with roots in the
Dutch Indies, part of second generation Dutch Indies
community
• Indisch Herinneringscentrum Bronbeek, Arnhem, the
Netherlands
• participated in a workshop on Internet History
Writing on Decolonization in Dutch Indies
• how they deal with their memories, e.g., sharing,
preserving how they participate in the community
64. Pilot 3:
Remembrance Community
• Design Rationale for CH
Communities Online:
• objective representation of objects in historical
context
• users exploring the meaning of collection objects
• community for sharing members perspectives
• Membership as a sense of belonging identification
• Influence of individuals in a group
• Fulfillment of needs in terms of benefits and rewards
• Shared emotional connections to stories
66. Dissemination at CH events
• Demonstrating Agora at MW2011, MW2012
• Events panel @DISH2010 and @CRESC
• Ignite Amsterdam (Marieke van Erp)
• Kom Je Ook and PhDO (Johan Oomen Lora Aroyo)
• CMN2010 (Chiel van den Akker)
• KNAW E-Humanities group (Susan Legêne)
Best poster @ DISH’09 and SIREN'11
67. Dissemination at CS events
• Organizing a series of events workshops
• DeRiVE (2011, 2012) @ISWC
• PATCH (2011, 2012) @UMAP, @MM
• Lora Aroyo
• Understanding of events (keynote iSemantics'2012)
• Events annotation (keynote SemWeb meetup NYC)
• Guus Schreiber
• Web Science: The Digital Heritage Case (keynote
SOFTSEM2010)
70. External cooperation
• Piek Vossen, Semantics of History
• Roxane Segers, PhD student Semantics of History
• Matje van de Camp, PhD student HiTime
• Yiling Lin, PhD studentUniversity of Pittsburg
71.
72. Roxane Segers, PhD student
Thesaurus of Historic Events
• Design and use of the Simple Event Model
• Journal of Web Semantics, 2011
• Extracting and Modeling Historical Events to Enhance
Searching and Browsing of Digital Cultural Heritage
Collections
• ESWC 2010
• Hacking history via event extraction
• K-CAP 2010
• Facilitating Non-expert Users of the KYOTO Platform
• LREC 2010
73. Johan Oomen, PhD student
Crowdsourcing user-generated content
• User participation in cultural heritage
• CT 2011
• Video Tagging: Waisda? Game
• WebSci 10, Ashgate chapter
• Oorlogsmonumenten crowdsourcing
• MW2011
• Content selection curation infrastructure
• Semantic Digital Archives 2011
• Open Content
74.
75. Implications for Education
• E-Humanities
• Computational approaches to Humanities problems
• What are the skill sets needed for a e-humanities
professional/researcher?
• Should we set up new types of interdisciplinary
programs?
• At bachelor/minor/master level?
76. Thank you for your
attention
http://agora.cs.vu.nl
@agoraproject