Crowd-sourcing the creation of “articles” within the Biodiversity
Heritage Library
Bianca [email protected]
Trish [email protected]
The BHL is…
• A consortium of 13 natural history, botanical libraries and research institutions
• An open access digital library for legacy biodiversity literature.
• An open data repository of taxonomic names and bibliographic information
• An increasingly global effort
BHLLITA 2011
Problem: Books vs. ArticlesLibrarians manage books Users need articles
BHLLITA 2011
Solution: “Article-ization”
Creating articles manually, through the help of our users: BHL PDF Generator
Creating articles through automated means: BioStor http://biostor.org/issn/0006-324X
BHLLITA 2011
Page, R. (2011). Extracting scientific articles from a large digital archive: BioStor and the Biodiversity Heritage Library. BMC Bioinformatics, 12(187). Retrieved from
http://www.biomedcentral.com/1471-2105/12/187
LITA 2011 BHL
Create-your-own PDF
BHLLITA 2011
What is an “article” anyway?
BHLLITA 2011
the Good, the Bad, the Ugly
BHLLITA 2011
the Good, the Bad, the Ugly
BHLLITA 2011
the Good, the Bad, the Ugly
BHLLITA 2011
Questions for Data Analysis
• What is the quality, or accuracy, of user provided metadata?
• What kinds of content are users creating?
• How can we improve the PDF generator interface?
BHLLITA 2011
Stats
• Jan 2010-Apr 2011 – Approx 60,000 pdfs created from PDF
Generator– 40% of those (approx 24,000) were ingested
into CiteBank (PDFs without user-contributed metadata excluded)
• 5 reviewers analyzed 945 pdfs (approx 3.9% of the 24,000+ articles going into Citebank)
**Thanks to reviewers Gilbert Borrego, Grace Costantino, and Sue Graves from the Smithsonian Institution
BHLLITA 2011
Methodological approach
• Quantitative – numerical rating system
• Rated titles, authors, beg/end pages• Its “findability” within CiteBank
search often determined how it was rated
BHLLITA 2011
Ratings System
Title
• 1=has all characters in title letter for letter• 2=does not have all characters in title letter for
letter but still findable in CiteBank search • 3= does not have all characters in title letter for
letter and is NOT findable via the CiteBank search
LITA 2011 BHL
Ratings System
Author
• 1=has all characters in author(s) last name letter for letter
• 2=has at least one author’s last name spelled correctly
• 3=has no authors or none of the author’s last names are spelled correctly
LITA 2011 BHL
Ratings System
Article beginning & ending pages
• 1=has all text pages for an article, from start to end
• 2=subset of pages from a larger article • 3=a set of pages where the intellectual content
has been compromised.
LITA 2011 BHL
Analysis steps
LITA 2011
ResultsTitle average
1.68
Title average 1.68
Author(s) average 1.33
Beg/End pg average 1.41
Title & Author average 1.50
Overall average (combines first 3 above)
1.47
LITA 2011 BHL
What did we learn?
• Ratings were better than we expected
• Many users took the time to create decent metadata
• “good enough” is not great but is still “findable”
LITA 2011 BHL
BHL-Australia’s new portalhttp://bhl.ala.org.au/
there’s always room for improvement
Other factors
But of course…..
BHLLITA 2011
Changes we madefor UI so far
• Asking users if they want to contribute their article to CiteBank
• Making article title a required field and validating it so its at least 2 or more characters
• Review button for users to review page selections and metadata (inspired by BHL-AUS)
• Reduced text and increased more intuitive graphics (inspired by BHL-AUS)
BHLLITA 2011
Brief survey of proposed changes
• Overwhelmingly positive response to proposed change
there’s always room for improvement
But of course…..
BHLLITA 2011
Success Factors
• Monitor the creation of the metadata to look at user behavior and patterns
• Engage with your users
• Incentivize your users
LITA 2011
@BioDivLibrary
/pages/Biodiversity-Heritage-Library/63547246565
/photos/biodivlibrary/sets/
/group/biodiversity-heritage-library
Bianca [email protected]
Trish [email protected]
http://biodiversitylibrary.org
BHLLITA 2011