CORE “Similar documents” widget

What with the Finch report making a bit of a splash right now and by all accounts downplaying the role of repositories and Green Open Access (expect a formal response from UKCoRR in due course, though in the meantime see this critical overview from Peter Suber – https://plus.google.com/109377556796183035206/posts/DsBAeSCofDX) it seems like a good time to remind ourselves of some of the exciting developments in repository world that should help us make Green a more viable alternative to Gold than the Finch report might seem to suggest. One such (technical) development is the Connecting Repositories (CORE) project at the Open University who have developed a widget that can be embedded in your local IR to generate links to similar documents from other repositories harvested by CORE.

The original CORE project focussed on harvesting papers from repositories and used text mining techniques to calculate ‘semantic similarity’ between different papers. This data was published openly, but not really in a format that could easily be used by others (at least outside a specialist community) so the CORE team have developed some simple javascript that, in theory, can be dropped into any repository setup and that will take the current document being displayed in the web interface and, if it is present in CORE data with some similar documents listed, will display links back to those documents in the CORE interface (these papers may come from any CORE harvested repository).

Obviously this means that CORE needs to be harvesting full text from your repository with a reasonable success rate for the plugin to work effectively (data on whether your repository is being harvested is available at http://core.kmi.open.ac.uk/repository_analytics/), although where there are no similar documents it should do nothing, so it won’t do any harm in any event.

The plugin is based on JQuery (jsonp call) which is compatible with any browser and any library system and can be styled to fit the UI of any repository. Currently ORO is the only repository to have implemented the plug-in (see this example record – http://oro.open.ac.uk/7123/) where it works only for full-text resources, though the team have developed an update that recommends full-text resources to metadata records (i.e. if there is a record in repository A that is metadata only, it will try to recommend related papers that do have full-text available), the update also filters duplicities and is more flexible.

The widget is currently available as a plug-in for EPrints from http://core-project.kmi.open.ac.uk/files/widget3.0.zip and the team are considering making it available from the Bazaar but are happy to assist in implementing in repositories running on other software.

Contacts are:

Owen Stephens (CORE Project Manager) – owen@owenstephens.com / @ostephens

Petr Knoth (Research Associate from the OU working on CORE) – p.knoth@open.ac.uk

Chris Yates (Systems Librarian at the OU who has implemented the plug-in in ORO) – c.s.yates@open.ac.uk / @chris_s_yates

Leave a Reply

Your email address will not be published. Required fields are marked *

Back To Top