Web Magazine for Information Professionals

Sharing History of Science and Medicine Gateway Metadata Using OAI-PMH

David Little outlines the resource sharing arrangements between the MedHist gateway and the Humbul hub, using the OAI Protocol for Metadata Harvesting, and some of the issues it has raised.

The MedHist gateway [1] was launched in August 2002, providing access to a searchable and browsable catalogue of high quality, evaluated history of medicine Internet resources. MedHist has been funded and developed by the Wellcome Library for the History and Understanding of Medicine [2], but is hosted by the BIOME health and life sciences hub [3], and as such is part of the Resource Discovery Network (RDN). MedHist was developed principally to fill the gaps left in the coverage of the history of medicine by existing resource discovery services within and outside the RDN. Both the Humbul Humanities Hub [4] and OMNI [5 gateway within BIOME provided some coverage of the subject, although this was not exhaustive. Outside the RDN, resource discovery services for the history of medicine were either defunct or concentrated on far narrower or broader subject areas [6] .

The fact the history of medicine is such an interdisciplinary subject caused problems for the Wellcome Library in deciding where to locate MedHist. Keen to keep the service within the RDN, it was decided to make MedHist a part of BIOME whose federated structure of health and medicine related gateways under a single hub suited the creation of an independent gateway with an affiliation to an existing service.

However, the interdisciplinary nature of the subject area suggested that it would also be important to make available MedHist's resource description records to other services with over-lapping subject interests, such as Humbul and the SOSIG [7] gateway, and to import any relevant metadata from other gateways, such as Humbul's History and Philosophy of Science records. Therefore, early on in the development of MedHist, methods of making its metadata available to other services, and importing external metadata, were investigated. The solution that was decided upon was the use of the Open Archives Initiative Protocol for Metadata Harvesting (OAI-PMH) [8].

Gateway metadata

MedHist, in line with other RDN gateways collects and makes available descriptive metadata about Internet resources, catalogued in accordance with the Dublin Core Element Set [9]. In addition to obvious access points such as title and URL, resource descriptions include an evaluative paragraph outlining the purpose and main features of the resource, and keywords are assigned from the National Library of Medicine's MeSH (Medical Subject Headings) thesaurus [10]. Additionally, where a resource is dedicated to an individual, personal name headings from the Library of Congress Name Authority File (LCNAF) are also added [11]. In addition to this descriptive metadata, administrative metadata such as site creator and owner are also collected, but not displayed on the MedHist Website.

Approaches to resource sharing

MedHist records automatically are available via a number of different access points: via the MedHist Website, the BIOME Website which searches the catalogues of all its constituent gateways and the RDN ResourceFinder database, a "union catalogue" of all the RDN's service providers' catalogues [12].

However, sharing metadata directly between different RDN services required the use of a separate process. Several options were considered before deciding on the use of OAI-PMH to expose and harvest metadata:

Using OAI-PMH to share metadata records

At present, MedHist records are exported to Humbul and Humbul's History and Philosophy of Science (HPS) records are imported into MedHist on a weekly basis. Whilst the exchange of records has been largely successful, some ongoing issues, some examined below, have currently prevented MedHist records being displayed on Humbul. However, Humbul HPS records are fully accessible via MedHist.


Importing Humbul HPS records into MedHist

Humbul OAI records are currently exported to MedHist on a weekly basis:

Once imported, MedHist staff:

Once live, Humbul HPS records are available on the MedHist Website:

Record display in Humbul

Record display for the Aldous Huxley: the author and his times Website

Record display in MedHist

...and the same record displaying in MedHist

Some issues

To date the process has highlighted some issues which are currently being addressed by both Humbul and MedHist.

Staff time overheads
To fully integrate Humbul records into MedHist, MedHist staff must spend time adding subject headings to imported records, and deleting any Humbul records which "duplicate" existing MedHist entries. In terms of the number of records imported from Humbul this is not too time consuming, although it raises questions of the sustainability of the process if it were extended to include records from other gateways. One approach currently being considered is the automated addition of suitable subject headings by the data provider, i.e. for Humbul to add agreed MeSH keywords to each of the records exported. This would allow records to be incorporated into MedHist's browse structure, although it is likely they would have to be very generic headings (e.g. "Science" and "Philosophy"). Whether it is possible, or even desirable, for third party metadata to be in an automatic "live" state after import is another area which needs to be examined more closely.

Re-presenting metadata
MedHist and Humbul have slightly different conventions for the display of metadata records. Humbul favour both a short and full record display, the latter displaying information such as site author and publisher in addition to title, URL and description etc. MedHist has only one record display which features hyperlinked title, description and keywords. BIOME have tended only to use author / publisher information only for internal purposes, whereas Humbul consider this part of the full record display, in the same way it would be displayed for books within a library OPAC. Certainly BIOME have expressed some concern about this administrative data being displayed and the implications it may have for data protection. This is an area which may need to be considered across the RDN as a whole.

In a similar vein, some thought has had to be given about the expression of rights statements when displaying third party metadata. At present MedHist displays a single rights statement which indicates the record is from Humbul's History and Philosophy of Science collection, but does not reflect the rights statements published by Humbul which acknowledge the work of individual cataloguers within its distributed cataloguing system. The most likely way that this will be addressed is to have a hyperlinked rights statement that will lead users to the full record within Humbul, where they will also be able to see the resource creator and publisher information.

Collection development
Over-reliance on third party metadata could potentially encourage gateways not to catalogue within certain areas of their collections which may be covered by other gateways. Whilst this may cut down duplication of effort and be seen as beneficial, it may be some key resources will not have been described by a particular gateway for its key audience. MedHist continues to catalogue all history of medicine resources, including ones that may already be within Humbul, and views other Humbul records as ways of supplying records which may be slightly more peripheral, although still of interest to the subject area. In addition to "adding value" to the service, these more tangential records help to provide contextual information for those interested in, for instance, developments within scientific thinking during a particular period of time.

OAI and Dublin Core
At present, OAI records only support the use of unqualified Dublin Core. This means that records imported and exported cannot express the full richness of the metadata collected. For instance, MedHist keywords are exported without any indication they are from the MeSH or LCNAF thesauri. Similarly, Humbul's author / creator distinctions (e.g. Web designer, author, compiler) are lost during the export process (although it must be noted that these would not be able to be used by MedHist in any case).

Different records for different services?
At present, BIOME must currently export two sets of OAI records: one for the RDN which feature a "cut-down" version of its gateways' metadata (featuring basic descriptive metadata and metadata about the metadata record itself), and one for Humbul which must additionally feature author and publisher information. This raises questions about standards within the RDN and the extra work for technical staff which "bilateral" agreements between gateways can create. This is an issue that will probably have to be considered across the RDN as a whole.


It has become clear that OAI is an effective way of sharing metadata between gateway services, but that it is not a panacea for all interoperability ills. The process between MedHist and Humbul has not been as straightforward as originally envisaged. It has been a "learning process" which has raised almost as many questions as it has solved. However, it is clear that the issues that have been raised are ones which may have to be addressed by anyone using OAI-PMH or metadata aggregation services and are hurdles to be overcome. Overall OAI-PMH has shown itself to be an efficient and effective way of metadata exchange, and also demonstrated how data may be re-used and re-formatted outside its original context. We certainly look forward to resolving some of the outstanding problems and pushing the resource sharing agenda forward with other related gateways, including SOSIG.


  1. MedHist is at: http://medhist.ac.uk
  2. The Wellcome Library home page is at: http://library.wellcome.ac.uk
  3. http://biome.ac.uk/
  4. http://www.humbul.ac.uk
  5. http://omni.ac.uk
  6. The Medical History on the Internet site: http://www.anes.uab.edu/medhist.htm closed as of December 2000. For a service concentrating on a narrower topic see the Karolinska Institute History of Biomedicine page: http://www.mic.ki.se/History.html; for a service covering a broader area, the ECHO history of science, technology and medicine pages: http://echo.gmu.edu/center
  7. http://sosig.ac.uk
  8. The Open Archives Initiative Website is at: http://www.openarchives.org
  9. The list of current Dublin Core elements can be accessed at: http://dublincore.org/usage/terms/dc/current-elements/
  10. The MeSH 2002 home page is at: http://www.nlm.nih.gov/mesh
  11. The Library of Congress Authorities page is at: http://authorities.loc.gov
  12. The RDN ResourceFinder catalogue is at: http://rdn.ac.uk
  13. See for instance Pete Stubely, 'Clumps as catalogues: virtual success or failure?', Ariadne Issue 22, http://www.ariadne.ac.uk/issue22/distributed/distukcat2.htm
  14. The My Humbul page is at: http://www.humbul.ac.uk/user/myhome.php
  15. More information on RSS can be found at: http://groups.yahoo.com/group/rss-dev/files/specification.html
  16. The RSSxpress page is at: http://rssxpress.ukoln.ac.uk
  17. See Pete Cliff, 'Building ResourceFinder', Ariadne Issue 30, http://www.ariadne.ac.uk/issue30/rdn-oai/

Author Details

 David Little
MedHist Project Officer
Wellcome Library for the History and Understanding of Medicine
Wellcome Trust

Email: d.little@wellcome.ac.uk
Web site: http://library.wellcome.ac.uk