Integrating DSpace publishing with Pubmed publishing

When your users publish in your local archive, they may certainly want Pubmed to be aware of their publication. Pubmed has different contributions procedures and XML metadata formats which could be directly generated from your DSpace repository (the inverse of the XSLT crosswalk we already defined). For instance...
Contributions to "LinkOut" PubMed service could also be integrated.

Improving DSpace user interface for big databases

The DSpace user interface seems very well suited for a hundred documents but it lacks many functionnalities to help the users "dig" in a big database. We are thinking of:

1. Making all indexed field displayed (authors, titles, subjects, dates) linked to searches on these fields. Little icons like those of Wikipedia could provide different search scopes: within the current search, within the current selection or within the whole database.
2. Quantities should be indicated at different places (number of records for an author or a subject, number of subjects for a letter or a prefix, etc.)

Suggestions are welcomed !

Importing or linking corresponding full-text documents

The preceding steps only create bibliographic records: getting a PDF copy of an article is very difficult to automate. We will be looking at ways to present the "LinkOut" display of the article and to easily upload manually a PDF when it is obtained from the author or elsewhere.
Care must be taken on "licensing": a PDF may be legally available to co-workers but rarely for free distribution on the Web. "Permalinks" to external documents are also an issue.
The advantage of loading a local PDF copy is full-text indexing...

Automatically importing Pubmed records in DSpace

Two programs are already developped:

1. An ingestion module which can import an XML file for which an XSLT crosswalk has been configured: this XML file can contain multiple bibliographic records which are then all loaded automatically.
2. An XSLT crosswalk from Pubmed XML format to an DSpace "overloaded DC" format.

This XSLT uses a Java function for Medline date conversion (we did not succeed to use JavaScript up to now): the corresponding compiled java class must therefore be included in the CLASSPATH.

When the development will be stabilised and documented, we will be happy to contribute it to DSpace community. Meanwhile, you can follow the links for an "as is" copy.

NLM Entrez is offering a collection of tools to automatically retrieve PubMed information.

A sample file is provided by NLM for testing purposes. The import result can be viewed on line.

Encouraging Pubmed selection and DSpace contribution

One of the main goal of this project is to have the "consumers" as "actors" of their choices: instead of proposing a selection in Pubmed, they are encouraged to select articles by themselves and to propose them to their colleagues. Users will have to learn how to use Entrez correctly and to save their selections in local files. An automated download and DSpace import is being developped to feed DSpace with users' selections.

DSpace metadata structure

Medline metadata structure is different from other bibliographic applications: history, specificities from Medicine and database size entailed most of the differences. Achieving a nearly lossless hosting of Pubmed data request to extend the standard DSpace "DC" metadata. A reasonable compromise has been defined for the Belgian PoisonCentre which adds:

1. contributor.Affiliation : Affiliation of first author
2. coverage.references : Number of references from the document to others
3. identifier.issue : Journal issue
4. identifier.volume : Journal volume
5. identifier.pagination : Journal issue pages interval (Medline format)
6. identifier.nlmuid : Journal NLM unique id.
7. identifier.pmid : Article Pubmed id.
8. publisher.Country : Country of publication
9. rights.abstract : Copyright on abstracts
10. subject.person : Person (or organisation) as a subject
11. subjetc.substance : Linked chemical substance (Future links included as [cas:...] or [einecs:...])

Existing fields which are used:

1. contributor.author : persons (lastname, firstnames) and organisations
2. date.issued : Publication date
3. description.abstract : Abstract(s)
4. identifier.citation : Medline "source" field (concatenation of journal abbreviation, publication date, volume, issue and page)
5. identifier.issn : ISSN of journal
6. language : three letters code of Article language
7. subject.mesh : Mesh subjects with qualifiers and "major topic" indicators
8. subject.other : Other keywords
9. title.alternative: Transliterrated title
10. title : English title
11. type : Publication type(s)

Pubmed Dates (other than PublicationDate) remain to be better managed.

Project steps

1. Defining a DSpace metadata structure to host Pubmed records
2. Encouraging Pubmed selection and DSpace repository contribution
3. Automatically importing Pubmed records in DSpace
4. Importing or linking corresponding full-text documents
5. Improving DSpace user interface for big databases (20 thousands records and more)
6. Integrating DSpace publishing with Pubmed publishing

Introduction

This blog is part of Destin inc. efforts to answer the Belgian Poison Centre needs for a bibliographic database and for knowledge sharing. We aim at federating the worldwide efforts to integrate PubMed (Medline) database (accessible through the Entrez user interface) with D-Space application.

Copyright and Disclaimer of NCBI for PubMed content and tools.