DSpace, ETDs, Automatic Metadata Extraction

1
DSpace, ETDs, Automatic Metadata Extraction

• Bradley Hemminger
• Jackson Fox
• Mao Ni
• School of Information and Library Science
• University of North Carolina at Chapel Hill

2
Big Picture

• Developing ETDs and Digital Libraries at UNC as
  part of a bigger vision. Use the development of
  ETDs at UNC to drive
• Digital Library support at UNC
• Advancement of DL and ETD research

3
Vision (NeoRef)

• Open Public Repositories of Knowledge
• Free, accessible anywhere 24/7/265
• Author self contribution
• Using templates that capture metadata
  automatically transfer metadata along with
  content
• Submission is point and click on file,
  verification of metadata, and selection of rights
  administration entire submission process should
  take less than 1 minute.
• Searching
• Simple and full metadata searching, full text
  (Google), with interactive drill down (Neoref).

4
Background

• Evaluation of ETDs in US.
• What sites are producing significant numbers of
  ETDs?
• What systems are in common use to support ETDs?
• How effective are these systems for the
  submitters, administrators, and searchers?

5
ETD Status Summary Oct 2003

• 17 active programs found via NDLTD
• Which application used?
• contentdm 1
• etd-db 11
• other 5
• Average of 1211 records
• Records available through the NDLTD union
  catalog?
• Available 4
• Not available 13
• Records available through NDLTD OAI catalog?
• Available 6
• Not available 11
• Total Records available through OAI 4995
  (average of 991 records)
• Contrast with today (Current List of OAI ETDs)

6
Contenders for DL for ETDs

• Virginia Tech ETD-DB
• EPrints
• DSpace
• ProQuest
• roll your own

7
Analysis of DSpace for ETDs

• Author contribution process long and painful.
• Limited rights administration (comes with single
  choice defined for MIT).
• Version 1.1 is not fully OAI compliant (fixed in
  1.2).
• Limited capabilities for showing metadata. When
  new metadata fields are created, there is no way
  to specify whether they should be displayed or
  not.

8
Analysis Continued

• Collection based metadata not supported (yet).
• Would help to have fewer assumptions built into
  the user interface.
• Contributor vs author confusion in reporting
  (dc.contributor.advisor listed as author as well
  as dc.contributor.author). Changed metadata used
  from dc.contributor.author to dc.creator.

9
Analysis Continued

• Add support for common metadata formats support
  is needed for exporting ETD-MS metadata
  (currently requires customizing the OAIcat
  software)
• Full text search not implemented yet
• Support for more flexibility in connecting local
  user identification/authentication schemes is
  desirable. DSpace assumes email IDs will be used.

10
Improvements made to DSpace

• Automated metadata extraction form author self
  contributed materials
• Rights administration extended to support
  Creative Commons
• User authentication based on UNC ONYEN

11
Simplify Input of Items
12
Workflow
13
Our starting point into DSpace
14
Rights Administration
15
Drawbacks to using DSpace

• Customization required. These cause problems
  later when you upgrade.
• Some important functions not available yet
  (collection based metadata, full text
  searching,).

16
Experience with SILS theses

• Customized version of DSpace used
• 70 masters papers submitted in spring 2004.
• Comments solicited. Most feedback came from
  students and staff who made themselves available
  to help students in the submission process.
• Comments provided by administrator as well.
• No feedback from searchers yet.

17
Comments..

• Problems encountered were mainly with using Word
  template (this was a new addition as part of ETD
  process), and conversion from Word to PDF.
• DSpace
• Users had to create a DSpace identity to submit,
  even though just for one use.
• After submitting users cannot edit their item
  until it has been reviewed by administrator.

18
Administrator comments

• Because of DSpace information flow, a student
  bringing a question to administrator about theses
  issue, could not view the thesis on the
  administrator screen, because it had been
  rejected (sent back) and was not longer
  accessible by them.
• When reviewing only abbreviated metadata list is
  seen (we changed default to show all metadata).
• More convenient and powerful interface for
  administrator would be helpful.
• Administrator needs to know specific item number
  in order to edit.
• Administrator has to re-enter through separate
  administrative interface in order to be able to
  edit items in their collection.

19
UNC Plan

• Locally use DSpace in Library to house and make
  available UNC produced theses and dissertations.
• At the same time, use ProQuest as off site long
  term archive for all materials, and bepress for
  submission process.
• Using SILS as prototype to develop and test
  effectiveness of DSpace for ETDs.

20
END
21
EXTRAS
22
(No Transcript)
23
(No Transcript)