Slides for presentation in session "Using LIDO in the Real World: Emerging Practice in Museum Metadata Sharing" at MCN 2014, the Museum Computer Network annual conference, in Dallas on 11/22/2014. #MCN2014 #MuseumCN
Introduction to Multilingual Retrieval Augmented Generation (RAG)
MCN 2014: Make One, Contribute Many: Sharing Image Metadata via LIDO
1. Using LIDO in the Real World:
Emerging Practice in Museum Metadata Sharing
In presentation order:
Erin Coburn, Independent Consultant
Emmanuelle Delmas-Glass, Yale Center for British Art
Rob Lance!eld, Davison Art Center, Wesleyan University
William Ying, Artstor
Piotr Adamczyk, Google Cultural Institute
MCN 2014, Dallas, 22 November 2014
2. Using LIDO in the Real World: Emerging Practice in Museum Metadata Sharing
Make One, Contribute Many:
Sharing Image Metadata via LIDO
Rob Lance!eld
Manager of Museum Information Services
Davison Art Center, Wesleyan University
@roblance!eld
MCN 2014
Museum Computer Network
Dallas
22 November 2014
3. Davison Art Center, Wesleyan University:
• More than 24,000 objects, chie"y
original prints and photographs
• Intensive access by students and faculty
for teaching and learning from objects
• Only 2.5 FTE total museum staff
(two full-time, one half-time)
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
4. With such a small museum staff…
• …time for anything is zero-sum with
time for many wholly unrelated things.
• …if the process that serves a new goal
is not every bit as efficient as it can be,
the time needed to serve that goal will
have undue impact on existing work,
seriously delay the new work, or both.
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
5. For DAC content sharing to gear up in
that context, we needed a means of…
• …minimizing time needed to prepare
structural aspects of metadata sharing
• …contributing to multiple aggregators
(and ideally, individuals online as well)
via one format, not tailored one-offs.
• …quickly generating new data sets.
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
6. The clear candidate: LIDO (Lightweight
Information Describing Objects).
• LIDO could carry all needed content.
• The main image aggregators we’re
working with could both accept it.
• Our collection information system had
the prospective capability to export it.
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
7. Our means of generating LIDO needed to…
• …leverage existing tools and resources.
• …enable rapid tuning of structure and
mapping: what !elds, what elements.
• …support not only batch contributions,
but also direct public download online.
DAC Collection Search infrastructure, built on
EmbARK Web Kiosk, helped us do this quickly.
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
8. Web Browser
LIDO implementation at the Davison Art Center, Wesleyan University
EmbARK Collection Manager collection
information:
full dataset
4D Server
Virtual Machine in University ITS Data Center
DAC Collection Search
page for an object record
LIDO link on
object page
DAC Collection Search: dac-collection.wesleyan.edu
EmbARK Web Kiosk
4D Web Server
DAC LIDO
template
(just an
XML file)
collection
information:
redacted
Virtual Machine in University ITS Data Center
2. The datafile
is periodically
copied for web
use. Before the
copy goes onto
the web server,
all content in
certain fields
is redacted (as
always, not only
because LIDO
will be served).
1. Back of house: no LIDO is generated here.
Along with general metadata, LIDO-specific
content is set here to later populate elements
such as lido:recordType, and to mesh with
EmbARK/4D conditionals in the DAC LIDO
XML template to support on-the-fly control
of such things as lido:resourceSet inclusion.
3. DAC Collection Search then uses
EmbARK Web Kiosk, the redacted
DAC datafile, and a DAC LIDO XML
template to create LIDO in response
to web browser requests over http.
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblancefield #MCN2014
9. XML template on DAC Collection Search server (runs on EmbARK Web Kiosk / 4D):
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
10. Template uses 4D HTML Tags to pull in content from EmbARK Web Kiosk data !le:
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
11. 4D conditionals control inclusion of certain LIDO elements for each object record:
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblancefield #MCN2014
13. Public-facing LIDO link (way down in the fine print) on each object-record page:
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblancefield #MCN2014
14. Excerpt from LIDO generated on the y from public LIDO link on an object page:
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
15. Conditional element for works with no known copyright restrictions:
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblancefield #MCN2014
16. Conditional resource elements for works that have DAC Open Access Images:
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
17. The template is just XML, making revision straightforward. Compare this part of it…
21. Anyone can download as many single
LIDO records as they wish. How to make
batch records for institutional supply of
single-!le metadata for image groups?
• Same XML template (only one to tune).
• Private, crafted URL makes LIDO !le for
a found set of records (private due to
its server load, not a policy constraint).
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
22. To minimize the ongoing/recurrent time
it takes to share DAC object metadata, this
approach focused on up-front tuning of
one automated way to produce LIDO.
But—does not tailoring metadata on the
museum (supply) side risk increasing the
work each aggregator needs to do? How
best to reduce the chance that it might?
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
23. THINK BIG:
To reduce the need for one-off parsing of
LIDO from different museums, aim to use
it as consistently as possible everywhere.
“But everybody’s LIDO is different…”
—An Anonymized Aggregator’s lament
We in museums may elicit this refrain less
often by making our LIDO more alike.
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
24. start small:
The Davison Art Center and the Yale
Center for British Art both have signi!cant
print holdings and use LIDO to contribute
metadata. We’ve started working to…
• …harmonize our LIDO practice as much
as we can in each of our local contexts.
• …document the differences necessary
to mesh well with local collection data.
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
25. (and aim to) Create…
1. …a better understanding of where
our current LIDO practices differ.
2. …a subtle evolution of those practices,
to harmonize them wherever possible.
3. …a “discordance” document, so tuning a
parser to ingest LIDO for prints from the
DAC and the YCBA may become easier.
4. …a LIDO pro!le for prints? Open question.
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
26. Starting with pairs of similar records, …
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
27. …find where they differ. For example:
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblancefield #MCN2014
28. A parting example of starting small, as illustrated in the preceding slide:
DAC and YCBA cataloging both include a !eld that
describes and transcribes inscriptions, combining
description and transcription for practical reasons.
LIDO has two elements in which that text might go:
inscriptionDescription and inscriptionTranscription.
A reasonable case can be made for or against each.
The DAC uses inscriptionDescription; the YCBA uses
inscriptionTranscription. Agreeing to use one of those
elements could incrementally simplify parsing LIDO
from the YCBA and DAC. Harmonizing practice in this
kind of way may make metadata aggregation easier.
Make One, Contribute Many: Sharing Image Metadata via LIDO | @roblance!eld #MCN2014
29. More about LIDO: http://lido-schema.org
Make One, Contribute Many: Sharing Image Metadata via LIDO
Rob Lance!eld, Manager of Museum Information Services
Davison Art Center, Wesleyan University | @roblance!eld
Using LIDO in the Real World: Emerging Practice in Museum Metadata Sharing
MCN 2014 | Museum Computer Network | Dallas | 22 November 2014
#MCN2014