Embed Size (px)
DESCRIPTION
Are the highly selective models of digital content creation satisfying user demands for increasing access to our vast collection holdings? In this era of decreasing library budgets and increasing responsibilities, is such a level of staffing possible at any but the well-funded libraries? As a recent article in the New York Times estimated, it would take 1,800 years for the National Archives to digitize its text holdings at the current rate of digitization1. Since November 2005, the University of Maryland libraries has engaged in another model for digitization: a workflow model that harnesses the digitization already being done by archivists and other staff for requests by patrons. By “crowdsourcing” selection decisions in this way the libraries have built a collection of over 5,000 objects from the holdings of the University Archives and Historical Manuscripts. This model is based on two main principles: · Selection: As one part of a programmatic approach to digitization, selections are based on user request and added to the publicly accessible digital repository · Image capture: Digitization itself proceeds on the premise that creating useful surrogates is more important than digital reformatting. The path to a successful workflow is fraught with perils, though. The presenters will discuss the issues that have proven most effective and most difficult in the large-scale digitization workflow in place at UM. They will highlight the technical requirements chosen for images, metadata, and quality control and speak about how they were, or in some cases were not, able to achieve them. In bringing to light these issues we hope to continue an ongoing conversation (most recently articulated at OCLC\'s \"Digitization Matters\" forum) about the purpose of digital collections and standards of digital surrogate creation, especially in the age of mass digitization projects. We hope to explore need to harness all of the library’s expertise and resources where they can best be deployed.
Citation preview
Crowdsourcing DigitizationHarnessing Workflows to
Increase Output
Gretchen Gueguen, East Carolina UniversityAnn Hanlon, Marquette University
LITA National Forum, 2008Cincinnati, Ohio
What is crowdsourcing?
• Jeff Howe, Wired Magazine, 2006– “distributed labor networks are using the Internet to exploit the
spare processing power of millions of human brains” – best example, Wikipedia…
– Any end achieved by harnessing the wisdom and labor of crowds
– Distributing the burden of a large endeavor
Howe, Jeff. “The Rise of Crowdsourcing”, Wired Magazine, Issue 14.06, June 2006
Crowdsourcing Digitization
• Crowd?– Patrons and Co-workers
• Capturing digitization for patron request– Selection is driven by patron request
• Centralized and Decentralized staffing for digitization
• Object: Build robust digital collections• Online collections dense enough for systematic
research (not just showcases)
Crowdsourcing Digitization
• The Wisdom of Crowds– How the project was conceived and developed: success story
• The Madness of Crowds– How the project failed, why: bringing it back from the brink
• Crowd Control– Methods used and lessons learned
• Attracting a Crowd– Critical mass for the masses: why we digitize
The Wisdom of Crowds
The Wisdom of Crowds
• Project Background: Archives and Special Collections– Digital image management for archives and special collections– Reducing redundancy – many items requested for digitization
more than once, why not track them?
• Digital Collections and Research (DCR)– New office to coordinate digitization efforts established– Establishing a digital repository – More ambitious than just image management
Image management = capturing patron scanning Image management = capturing patron scanning workflow to populate the new repositoryworkflow to populate the new repository
The Wisdom of Crowds
• Coordination between Archives and Digital Collections: – New metadata schema
– New best practice guidelines
• Developing Repository– Fedora required development
Meanwhile, patron scanning continues to grow…Meanwhile, patron scanning continues to grow…
The Wisdom of Crowds
• Answer: Scanning Database– Microsoft Access database: “stop-gap measure” while digital
repository was being built
– Corresponded to newly created XML schema and metadata requirements for repository
The Wisdom of Crowds
The Wisdom of Crowds
• Biggest beneficiary: University Archives– Receives the most scanning requests from patrons– Capture patron requests, as well as items scanned prior to
implementation of Scanning Database– University celebrating 150th anniversary
• Documentary• “Coffee table” book• Departmental histories• Nostalgic alumnae
The Wisdom of Crowds
• Collections created by crowdsourcing digitization:– University AlbUM– National Trust for Historic Preservation Postcard Collection
The Madness of Crowds
The Madness of Crowds
• Evolution– Evolving standards for both metadata and imaging
• Training/Quality• (dis)Organization• Backlog
www.funnyfreepics.com
The Madness of Crowds
• Evolution– Quality of legacy scans
• file types• spatial resolutions• Color profiles• Clipping, noise, and other
“problems”• Flawed equipment
• Training/Procedures• (dis)Organization
• Backlog
The Madness of CrowdsRotated 90º
Rotated 180º
24-bit color 300 dpi tif
8-bit 600 dpi tif
48-bit color 600 dpi tif
BitonalEPS
16-bit 300 dpi JPEG
indexed color 72 dpi gif
PDF???
The Madness of Crowds
• Evolution– Metadata Quality
• Lack of experience with controlled vocabularies and input standards
• Changing metadata requirements
• Training/Procedures• (dis)Organization• Backlog
The Madness of Crowds
It’s not quite wrong…
but, it’s not quite right
• Evolution
• Training/Procedures– No standard guidelines for scanning procedures
– No quality control procedures for images or metadata
– No one to set them up anyway…• (dis)Organization• Backlog
The Madness of Crowds
The Madness of Crowds
The Madness of Crowds
The Madness of Crowds
• Evolution
• Training/Procedures
• (dis)Organization― Does everything fit in
a “collection?• Backlog
The Madness of Crowds
• Evolution• Training/Procedures• (dis)Organization
• Backlog– Robust metadata standard to enable repurposing and “sharability”
– Could take 10x more time to do metadata than scanning
– Volume of scanning didn’t leave much time for metadata
The Madness of Crowds
Crowd Control
1. Create Documentation2. “Teachable” standard3. Responsibility4. Quality5. Divide and Conquer?!?
Crowd Control
Crowd Control
1. Create Documentation
2. TEACH it3. Responsibility4. Quality: Live it, Learn it, Love it 5. Divide and
Conquer
6. file format
3. straightness and placement 1.
resolution 2. color
4. reference points (targets)
5. noise
Imaging EnvironmentDefined
Image StateRAW
Prepped for a specific
output
Output Referred -
looks towards output
Input Referred -
looks towards sensorOriginal Referred - defined
relationship between original and digital version
Current Practice
Emerging Practice
More technical
metadata is needed
Should be able to get
by with less
technical metadata
Puglia, 2007
Crowd Control
1. Create documentation2. TEACH it!
3. Quality: Live it, Learn it, Love it– Have curatorial staff check for accuracy and completeness
– DCR staff follow up with a check of a statistically significant portion for style and consistency
– Eventually, give curatorial staff to make corrections as they find them using the web-based administrative form
4. Responsibility5. Divide and conquer?!?
Crowd Control
1. Documentation2. “Teachable” standard3. Quality: Live it, Learn it, Love it
4. Responsibility– Someone has to have some
– But it doesn’t have to be an entire job5. Divide and Conquer
Crowd Control
1. Create documentation2. TEACH it!3. Quality: Live it, Learn it, Love it4. Responsibility
5. Divide and conquer?!?– Stub record created at request time; Cataloging enhances
Crowd Control
Crowd Control
1. Create documentation2. TEACH it!3. Quality: Live it, Learn it, Love it4. Responsibility5. Divide and conquer
6. Give up• Less control, more power
Crowd Control
• Would you want to try this?– Give yourself room to evolve and change through the project– Don’t feel like every image is a precious snowflake– More than any single technique, it’s the philosophy of crowdsourcing
that’s more important
l
Access to a low-quality scan…
…is still better than no access at all.
Crowd Control
• Would you want to try this?
– Don’t feel like every image is a precious snowflake
• Would you want to try this?
– More than any single technique, it’s the philosophy of crowdsourcing that’s important
Crowd Control
Attracting a Crowd
Attracting a Crowd
• Letting Go– “Letting go” creates efficiencies– Looking at expertise across the Libraries– Distribute the burden
Move away from “trophy” collections Move away from “trophy” collections
toward online Research Collectionstoward online Research Collections
Attracting a Crowd
• Distributed Problem-solving– Ideas from Archives:
• Organizing repository by subject rather than by collection• Dabbling in folder-level description (and digitization) rather
than just item-level
• Neutral Collection-building
Erway, Ricky and Jennifer Schaffner. 2007, “Gearing Up to Get Into the Flow.” Report produced by OCLC
Programs and Research (formerly RLG)
Attracting a Crowd
• Distributed Problem-solving– Ideas from Archives:
• Using “stub records” from patron request forms• Dabbling in folder-level description (and digitization) rather
than just item-level
• “Neutral” Collection-building― Wikipedia-style collection-building― Building a collection with wide
range
Attracting a Crowd
• Mass digitization– Google projects:
• Books• Newspapers
• Mass decision- making– Instead of item-level
decision-making
Attracting a Crowd
• Making Digitization a Core Function of the Library– Mission Statements come to life!– Organizing around digitization – very little has really
been done yet
Why? For researchersFor researchers
• “Fringe activities” need to become core investments
― Metadata creation― Digitization
Council on Library and Information Resources (CLIR). No Brief Candle: Reconceiving Research Libraries for the 21st Century, 2008.
Crowdsourcing Digitization
THANKS!
Access these slides at:http://www.personal.ecu.edu/presentations/Crowdsourcing.ppt
Or:http://www.slideshare.net
Gretchen [email protected] Carolina UniversityGreenville, North Carolina
Ann [email protected] UniversityMilwaukee, Wisconsin
