Socially inept

1
Socially inept?

• Shoehorning tags on a library

2
The Set Up

• At Georgia Tech we are toying with this. We are
  trying to create an alternative interface to the
  OPAC (exporting all the bib records as marcxml -
  most likely to be transformed to MODS) and trying
  to create a natual language search on top of it.
  Included in this would be folksonomic tagging (as
  well as authoritative LCSH) which could be
  included in the search (but possibly weighted
  less than authoritative taxonomies - that will be
  worked out later). Once enough items have been
  tagged, we will have the ability to map between
  folksonomic and LCSH (hopefully) so a user could
  conceivably (using WAG the Dog or some other
  form) move back and forth between folksonomic
  interfaces (del.icio.us, Furl, Flickr, unalog -
  although possibly more likely CiteULike and
  Connotea) and the OPAC and local resources.
• I figure that by allowing users to save certain
  items and searches with their own taxonomies,
  they will also be (inadverdently) opening up the
  catalog to other discovery methods. Win
  win,really.
• The net can be cast even wider by utilizing this
  with some sort of metasearch.
• --Ross Singer, Web4Lib April 4th, 2005

3
Charlatan!
4
So what happened?

• Enabling tagging, technically speaking, is
  relatively easy
• Enabling tagging, socially speaking, is not
• What have you got that a larger community
  doesnt? (hint, your OPAC isnt a huge sell)
• Just because you build it doesnt guarantee
  anybody actually coming

5
In the meantime weve seen

• CiteULike
• Yale Links
• PennTags
• Connotea
• LibraryThing
• Amazon Tags

6
A little about each of these
7
CiteULike

• Created by Richard Cameron in Nov. 2004
• Intended to bookmark and tag peer-reviewed
  scholarly journal articles
• Items can be added via bookmarklet (for certain
  publishers/aggregators) or uploaded from
  BibTex/EndNote/RefWorks/etc.
• Active community - 1000 posts/day
• Group support

8
Yale Links

• Developed by Dan Chudnov, launched September 2004
• Little to no institutional promotion
• Based on unalog
• Support for groups
• No real integration with library resources (OPAC)
• Service will most likely be ended soon

9
PennTags

• Developed by Michael Winkler and launched in 2005
• Seemingly strong institutional support
• Relatively small (yet relatively active) user
  base - 425 users (20k FTE), 20-40 posts/day
  (summer)
• Tags and annotations appear in OPAC
• Projects

10
Connotea

• Created by Nature Publishing Group
• Similar to CiteULike in scope/functionality
• No group functionality
• Similar numbers to CiteULike
• OpenURL Support
• API

11
LibraryThing

• Allows users to catalog their personal
  collections
• gt3M records (900K unique), 46.5K users
• Grabs metadata from Amazon, LOC 45 other
  libraries via Z39.50
• Recently released an xISBN-like service based on
  its data

12
Amazon

• Recently added tagging
• Gives Amazon yet another avenue to hawk their
  wares at consumers
• Tags quality wildly vary - editorial comments,
  etc.
• Sole purpose seems to be to promote more sales

13
So back to why Im here

• Social bookmarking support added to the ümlaut --
  an OpenURL Link Resolver service
• For a given citation, queries PubMed, SFX, local
  catalog, consortial catalog, Amazon, OAI
  Providers and stores any subject descriptions
  gleaned from sources
• Checks the fulltext target urls against Connotea,
  unalog and Yahoos MyWeb2.0 (more planned when
  APIs become available)
• Stores tags and tagcount

14
LCSH, MeSH, and tags, oh my!

• All terms given same weight -- LCSH
  headings/subheadings divided up
• More matching tokens greater relevance
  recommender service
• No browsing on a single term
• (Hopefully) this will create a map between
  LCSH/MeSH/OAI Subject Keywords/SFX
  Categories/tags to guess appropriate subjects for
  citations without matches in all areas

15
But this about Authority Control

• The problems with using tagging sources
• URLs are a very weak identifier of an item
• Items are only tagged (or should only be tagged)
  when they are found useful and wish to be found
  again
• The quality and focus of the source is the key
• The larger the data/user pool the more useful the
  data

16
To put some criticism in perspective

• Rob Sanderson - University of Liverpool
• Datamined MELVYL (export from several years ago)
  -- roughly 18M records
• Found 127,896 terms in the 650a
• Only 47,319 appear more than 5 times

17
Hybrid Moment

• Imperfect solution -- yet no other hope to relate
  structured data to the wild and woolly web
• Sites like LibraryThing, CiteULike and Connotea
  could aid building vocabularies with identifiable
  data
• Journal records sum of article tags?
• Authority may take the form of associating the
  proper community to a given object

18
Thanks, yall!

• Ross Singer
• ross.singer_at_library.gatech.edu
• http//dilettantes.code4lib.org/