Data and Publication Discovery

1
Data and Publication Discovery

• Brian Matthews,
• Information Management Group,
• STFC Rutherford Appleton Laboratory
• CLADDIER workshop, Chilworth, Southampton, UK
• 15th May 2007

2
Microsofts Science 2020 Report

• Modern scientific communication relies on both
  journals and databases. At present these are not
  integrated.
• By 2020 mutual linking will be commonplace and
  publications just containing peer-reviewed data
  will become available.
• http//research.microsoft.com/towards2020science/d
  ownloads.htm

3
The Use Case

• Joanna, at the University of Southampton, has
  done some work on the biology of seawater off the
  coast of Cornwall. As part of her analysis she
  needs (from a number of locations)
• Publications and data describing prior or similar
  work.
• Oceanic profiles of salinity and temperature from
  the closest cruise in time and space,
• Meteorological data to accompany both her own
  sampling and the oceanic data,
• Remotely sensed ocean colour imagery (to add
  additional information on the biota).
• She will then publish a paper that cites the
  datasets, lodge the paper in her own
  institutional repository and also deposit her
  datasets in one or more appropriate data
  repositories (e.g. both the NOCS data archive,
  and the, BODC).
• The work Joanna has done is of interest in
  calibrating a global earth system model to
  compare simulations of oceanic CO2 production
  with the scenarios used in the model.
• Fred, at Reading University needs to be able to
  find Joannas paper and data either via citations
  or directly from publication repositories. Having
  found the paper, the data should be obtainable
  via the citation and the data archive.
• As part of his work he checks back through the
  other datasets used and cited as inputs to
  Joannas data, as before he uses Joannas data,
  he suspects Joannas work could be recalibrated
  by using better quality meteorological
  re-analyses.

4
(No Transcript)
5
What does that need?

• Joannas own data acquisition
• Location and acquisition of prior publications
  and data
• Location and acquisition of remote datasets
  required as part of the analysis
• Creation of personal metadata for new data
• Data analysis and paper writing
• Citation of remote papers and datasets
• Paper submission to a journal and acceptance
• Repository submission of paper (maybe a preprint)
• Repository submission of data
• Further metadata creation for the data (at the
  data repository).
• Further metadata creation for the publication (at
  the institutional repository)
• Linking between institutional repositories and
  the data held at the discipline repository
• All the datasets and publications cited need to
  be annotated with the citation information
• Discovery of Joannas work by Fred (either from
  Joannas publication or datasets or citations
  thereof)
• Acquisition of all the relevant publications and
  datasets by Fred
• Analysis and Publication by Fred (and all the
  same steps from 5 as required by Joanna)
• External Adjudicators need to be able to find and
  acquire citation information.

6
So what services do we need?

• In order to achieve this scenario we need to
  provide a set of key services
• Publishing of Data
• Browsing and searching
• across different repositories
• across data and publication
• Cross-citation of data and publication
• forward and backward citation
• need to maintain currency of citation links

7
Browsing and Searching

• Browsing and searching
• across different repositories
• across data and publication
• CLADDIER has provided a harvesting and search
  tool to support cross-repository searching
• Uses OAI-PMH a conventional approach
• Simple but it works!
• Simple key-word searching
• Three participating repositories in the pilot
• BADC, STFC ePubs, e-Prints Soton

8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
Adding cross-citation

• The Discovery Service gives a broad-brush search
• Give you both publications and data sets
• which are indexed on a key word
• A Google across repositories
• Currently, cannot tell whether the data and
  publication are actually related
• what data and publications inspire a piece of
  work (generating a new data set)
• what publications arise from a data set
• We need to exploit the concept of citation to see
  whether relationships are actually related

15
Traditional Citations
16
Cross-citation
17
Adding Citations to the Metadata Model

• Adding Citations has been considered in standard
  metadata models.
• e.g. Scholarly Works Application Profile
• JISC funded initiative
• Dublin Core Application Profile
• Describing Scholarly Publications (ePrints)
• Based on the FRBR model
• Does consider Citations
• But breaks citations up into small components
• This is highly labour intensive to enter
• Does not have a notion of back citation

18
FRBR Model
19
ePubs and Cross-Citations

• STFC ePubs has a metadata model based on FRBR
• Need to extend this to support cross-citation
• Keep it simple
• Can support forward and back links
• Have developed a simple model for citations

20
Citation Model
21
(No Transcript)
22
(No Transcript)
23
Maintaining Links

• Ideally the archives holding the datasets and
  publications would be notified that a paper
  citing them had been submitted.
• Metadata associated with those records would be
  updated to reflect the citations.
• The metadata in the publication repository should
  also link to the data in the data archives and
  vice versa.
• It would be great if this notification could be
  done automatically.

24
Notification Services

• To support this, we need to provide a
  notification service.
• Federated Repositories register with the service
• Repositories notify the service of citations
• The service informs (via broadcasting or
  targeting) repositories of citation,
• Service provides sufficient information to update
  metadata
• Still under development.
• Note Blogging software.

25
Conclusions

• The Use Case supports the scientific process
  with repositories
• This requires the cross-linking network of
  information objects
• Which needs to be stored, maintained and searched
• Tools and ideas relatively straightforward
• Lots of gluing of existing components
• Keep it simple so it will get used