A presentation by Andrew Greg, University of Glasgow. Invited talk at a workshop for the 'Scotland's Collections and the Digital Humanities' knowledge-exchange project, hosted at the University of Edinburgh. 12 September 2014. http://www.blogs.hss.ed.ac.uk/archives-now/
Engaging the public in tagging and researching the UK's paintings: Two case studies
1. Engaging the public in tagging and researching
the UK’s paintings: two case studies
Andrew Greg,
Director, National Inventory Research Project, University of Glasgow
andrew.greg@glasgow.ac.uk
Willem van Haecht
(1593-1637),
The Gallery of Cornelis
van der Geest, 1628,
oil on panel, 100 x 130
cm, Rubenshuis,
Antwerp
2. The projects
• National Inventory Research Project:
NICE Paintings
http://www.vads.ac.uk/collections/NIRP/i
ndex.php
• The Public Catalogue Foundation: Your
Paintings
http://www.bbc.co.uk/arts/yourpaintings/
• Your Paintings Tagger
http://tagger.thepcf.org.uk/
• Art Detective
http://www.thepcf.org.uk/artdetective/
3. National Inventory Research Project
NICE Paintings
• National Inventory Research Project (NIRP) set up
2001 to aid research in regional museums and
improve access to collection information
• Initiated by the National Gallery, London, and a
steering committee – now the Advisory
Committee for Research on European Paintings
• Now based in University of Glasgow, NIRP’s first
project 2004-7: 8,000 paintings in 200 UK
collections
• 2009 work with the V&A, the Royal Academy and
Apsley House
• Over 9,000 paintings and newly researched
records now online as NICE Paintings: the
National Inventory of Continental European
Paintings
• New project 2014-15 NIRP in the North, working
with four major collections: Liverpool,
Manchester, Leeds and York.
32 project researchers
and Neil MacGregor
Scholars (below, with Neil
MacGregor and Dr Susan
Foister of NG), 12 now in
curatorial posts, 7 in
academic art history posts
and 3 freelance art
historians.
4. Core record:
David Teniers II,
Interior of a
Tavern, oil on
panel, Shipley Art
Gallery,
Gateshead.
http://www.vads.a
c.uk/large.php?uid
=84081&sos=1
5. Full record:
David Teniers II,
Interior of a
Tavern, oil on
panel, Shipley Art
Gallery,
Gateshead.
http://www.vads.a
c.uk/large.php?uid
=84081&sos=1
6. The Public Catalogue Foundation
Your Paintings
Your Paintings comprises
images and basic
information on all 211,000
oil paintings in 3,000 UK
collections ranging from
hospitals to national
museums. Originally in book
form.
Data collected by PCF:
• Artist name and
attribution qualifier
(‘after’, ‘attributed to’,
etc.)
• Title
• Production date, if known
• Medium
• Size
• Accession number
• Acquisition method
Above: Public Catalogue Foundation, Paintings in
Public Ownership. West Yorkshire: Leeds, 2004, p.31
7. Your Paintings
now on line
hosted by BBC.
• Limited
information,
but
• Links to other
works by the
artist
• To other
works in the
collection
• To more
detailed NICE
Paintings
data and
• To Art
Detective.
Right: painting page
from Art Detective
8. Additional data
requirements for
effective searching:
• Structured content
keywords
• Names of people,
places and events
• Types, or genres of
paintings
• Dates and periods
• Styles and artistic
movements
• Social tags
The tasks and
constraints:
• 200,000 paintings
• respond to user
research
• reliable and
consistent
• minimise cost
• short timescale
9. Two examples of other
museums collecting public
tags:
1. Detail of object page
from Brooklyn Museum
with ‘Posse’ tags on the
right.
http://www.brooklynmuseum.org/
opencollection/objects/1452/Kiss_
Me_and_Youll_Kiss_the_Lasses
10. 2. Detail of object page from Indianapolis Museum of Art using the ‘Steve Tagger’ with tag cloud.
[N.B. Accessed 10 Jan 2012. New page accessed 8 Oct 2012 has a different layout with the tags at the bottom.]
12. Your Paintings Tagger
Search for a crowdsourcing
model
The model, Galaxy Zoo:
• Launched July 2007
• 1,000,000 images of
galaxies
• 150,000 worldwide
volunteers in first year
• 50,000,000 classifications
in first year.
Key principles:
• No expertise required
• Images delivered at
random
• Multiple classifications.
Results as good as
professionals.
www.galaxyzoo.org
13. Your Paintings Tagger
Workflows
Public workflows
• Social tagging (‘things and ideas’)
• Types
• Subjects
• Names (people, places and events)
Expert workflows
• Dates
• Styles and movements
Quality assurance
Thresholds
• 15 taggers per painting
• Tags accepted after:
– Things: 2 or more taggers use a term
– Names: 2 or more taggers use a name
– Types: 4 or more taggers select a type
– Subjects: 2 or more taggers select a
subject
Referrals to ‘Supervisors Interface’
– Specific problem subjects
– Two or more names selected in one
painting
– Abstracts, portraits and still lifes
14.
15.
16.
17. Launched in June 2011.
As of March 2013:
• 9,090 registered taggers of
whom 671 (7.3%) are ‘expert’
taggers
• 40% have tagged fewer than five
paintings
• Only 15.5% have tagged more
than 50 paintings
• 49 of these (0.5% of total) have
tagged over 1,000 paintings
each
• Over 3.5 million tags have been
created
• Over 23,000 paintings have been
completed.
Productivity of taggers March 2013
25.7%
17.3%
13.0%
41.5%
1.4%
0.6%
0.5%
No pictures tagged
1-4 pictures tagged
5-49 pictures tagged
50-249 pictures tagged
250-499 pictures tagged
500-999 pictures tagged
1,000 or more pictures tagged
Your Paintings Tagger
18. Your Paintings Tagger
Screenshots from Your Paintings Tagger, 27 August 2014
Supertaggers:
19. Your Paintings Tagger
• Tags for 23,000 paintings
delivered to BBC
• Now about 50-100
taggers active every day
• The first 2,278 completed
paintings received an
average of:
– 61 different free text tags,
of which 22 were accepted
– 15 different subject terms
(excluding ‘Other’), of
which 7 were accepted
– 87 different tags and
keywords of which 30
were accepted.
Right: detail of painting page on Your Paintings
showing 36 public tags for: Emily Mary Osborn
(1828-1925), Study for 'Nameless and Friendless',
1857, oil on panel, York Museums Trust
20. Your Paintings Tagger
Work with Oxford Visual Geometry
Group, which is developing image
recognition software.
Abstract:
“The objective of this work is to find
objects in paintings by learning
object-category classifiers from
available sources of natural images.
Finding such objects is of much
benefit to the art history community
as well as being a challenging
problem in large-scale retrieval and
domain adaptation…”
Image from: Elliot J. Crowley and Andrew
Zisserman, Visual Geometry Group, Department
of Engineering Science, University of Oxford, ‘In
Search of Art’, Workshop on Computer Vision for
Art Analysis, ECCV, 2014
http://www.robots.ox.ac.uk/~vgg/publications/2
014/Crowley14a/crowley14a.pdf
21. Your Paintings Tagger
Benefit of automated
tag creation for Your
Paintings Tagger,
through manual
check:
1. Taggers shown
computer-selected
images
and confirm
identification
2. In two months
250 taggers
confirmed nearly
1 million tags for
over 200
subjects in
99,600 paintings
22. Art Detective
Aims:
• Improve the knowledge
collection managers have
about the art in their care
• Improve communication
between curators and
sources of expertise
• Engage the public in
discussions around art
history, historical research
and connoisseurship.
A free-to-use online interface
bringing together:
• Curators in search of
specialist information
regarding their collection
• Specialist knowledge from
academics, the art trade
and other experts
• Interested members of the
general public.
Above: screenshot of early design of home page for Art Detective
23. Art Detective
• Anyone can browse
Art Detective
without signing in
• To make a
comment they
must sign up with
name and email
address
• To start a discussion
they must enter via
a specific painting
page on Your
Paintings (right).
24. Art
Detective
• Discussion topics on
Art Detective are
linked to special
interest groups with
expert group leaders
• Users can subscribe
to groups and
receive updates on
new discussions in
areas that interest
them
• Users can follow
particular
discussions.
Right: some of the regional and
thematic groups on Art Detective
25. Art
Detective
Screenshot of Art
Detective Discussions
page with most recently
commented upon
discussions from top
July 2014:
5176 unique users
(average time on site 4.5
mins)
47% returning visitors
(average time on the site
8 mins)
845 registered users (i.e.
contributors)
27. Art Detective
Some successes: identifications of a doctor through archival and newspaper research; and of some
back gardens through local knowledge.
28. Art Detective
And two outstanding mysteries, several months old: the identity of an obscure subject, possibly
connected to the Venetian poetess Veronica Franco; and the identity of a minor artist, possibly the
Dutchman Ary de Vois (c 1632-1680) or the obscure English gentleman artist Henry Gibbs (1631-1713).
29. Engaging the public in tagging and
researching the UK’s paintings: two
case studies
Thank you
• National Inventory Research Project: NICE Paintings
http://www.vads.ac.uk/collections/NIRP/index.php
• The Public Catalogue Foundation: Your Paintings http://www.bbc.co.uk/arts/yourpaintings/
• Your Paintings Tagger http://tagger.thepcf.org.uk/
• Art Detective http://www.thepcf.org.uk/artdetective/
Andrew.greg@glasgow.ac.uk