Social Navigation, Recommender Systems, and Libraries

1
Social Navigation, Recommender Systems, and
Libraries Kornelia Tancheva Jesse Koennecke
2
General Principles of Social Navigation

• Architecture, urban design, sociology
• Follow other peoples trails in space
• Deliberate or spontaneous
• Safe
• Space vs. Place

3
General Principles of Social Navigation
4
Short History and Typology of Social Navigation

• PERSONA Collaborative Project
• 1998 Workshop on personal and Social navigation
  of Information Spaces
• 1999 panel on SN at CHI conference
• 2000 panel at CHI
• 2001 Delos-NSF workshop on personalization and
  recommender systems (Dublin)
• 2003 Social Navigation of Information Space (2nd
  ed.)
• Direct and indirect

5
Recommender Systems

• Assist and augment the natural social process
• E-commerce and entertainment
• University of Karsruhe RS (2002)
• Melvyn Project (U of California)
• Collaborative Filtering
• Pull-active
• Push-active
• Automated CF
• Differences with non-virtual SN
• Methods
• Ranking
• Annotations, reviews
• Content analysis

6
Social Navigation and Libraries

• Nature of research domains vs. entertainment
• Research is not social?
• Personal vs. general
• Expertise vs. the wisdom of crowds
• Philosophical problems
• Information is not an objective set of data
• Interpretation subjective view of the world
• Cultural differences
• Technical problems
• Unstructured user-contributed data vs. structured
  legacy data
• Privacy
• Limitations of RS

7
Cornells Project

• General interest amongst library staff
• Call for Masters projects Fall 2005
• Three graduate students
• Cross disciplinary library staff
• Guiding principles
• Our materials, our patrons
• Patron privacy

8
The Data

• Patron Tags
• Undergraduates College
• Staff - Department
• Faculty - Department
• Graduates Field of Study or College
• Book Tags
• HILCC Hierarchical Interface to Library of
  Congress Classification

Davis, S.P. "HILCC, A Hierarchical Interface to
Library of Congress Classification "Journal of
Internet Cataloging, v.5, no 4 (2002), p.
19-49. HILCC at Columbia - http//www.columbia.ed
u/cu/libraries/inside/projects/metadata/hilcc/
9
Results

• Ranking dB Most circulated books per HILCC
  category
• Less than 10 circulated more than twice in five
  years
• Most level-1 HILCC categories had significant top
  20 lists
• Most level-2 and beyond did not
• Relation dB People who checked out this book
  also checked out these books
• Matching different the bibliographic records of
  books that were checked out to the same patron on
  the same day
• Connections that could be made
• Graduate students in Field of Study X who checked
  out this book also checked out the following
  books
• You may find other useful books in the following
  categories

10
What we learned

• We need more and better data
• Limiting to graduate students severely limited
  our quantity of data
• The ranking dB had a relatively small number of
  frequently circulated books
• The relation dB had very few instances of books
  being connected more than once
• We need to rethink our approach

11
Next Steps

• Access more descriptive data about our patrons,
  particularly our undergraduates, to expand our
  scope to more of our constituents.
• Collect our data in a more effective way to find
  more connections between people and books.
• Approach peer institutions to increase the amount
  of relevant data. This would also increase the
  potential user base for the RS.
• Include the ability for users to add their own
  descriptive tags and ratings. This would
  increase the total amount of usable data while
  adding some sense of ownership for the users.

12
Questions?

• Kornelia Tancheva
• Director, Collections, Reference, Instruction,
  and Outreach
• Public Services and Assessment - Cornell
  University Library
• kt18_at_cornell.edu
• Jesse Koennecke
• Head, Access Services
• Albert R. Mann Library
• jtk1_at_cornell.edu