The HRU (Tudhope and Binding) collaborated in the PERsonalised TAgging Information Services (PERTAINS) project, funded as part of the Joint Information Systems Committee (JISC)
is part of the JISC Personlisation Programme. The project was led by MIMAS and in investigated the integration of pilot tag recommenders in UK national services (Intute and COPAC).
PERTAINS aims to exploit Intute’s and Copac’s UK presence as powerful aggregators of learning and research materials by implementing a demonstrator tag suggestion system. We believe that the demonstrators developed could provide the foundations for a new tag recommender system and infrastructure for future development work in other services.
The project was a 5 month collaboration during 2009 between the University of Glamorgan and Mimas, University of Manchester, with funding for 50% developer time at each partner.
A tag recommender service was developed at Glamorgan and two client user interfaces employing the services for Intute and for COPAC were developed by MIMAS. The suggestion service was based on a SKOS represented vocabulary. The user interface included various suggestion options. This approach appeared fast, scalable, and platform neutral and the URL-based web service approach facilitated rapid development of the client user interfaces.
Two options were identified for PERTAINS services from the initial project meeting.
1.An initial (default) set of suggestions before the user enters a tag (based on document title, existing metadata, etc).
2.A suggestion made after user tag entered based on the user tag (plus using available ‘context’ from document title and metadata to help disambiguate between options)
A further option is an ‘area of interest filter’ which allows the user to refine suggested tags by selecting (multiple) options from a broad classification (in the case of DDC, the top three levels). The options correspond to (sub)hierarchies in the vocabulary. This area of interest can be employed to reduce the number of possible suggestions when a very general tag is entered.
A URL-based API was developed at Glamorgan for the bespoke PERTAINS vocabulary based recommendation services. JSON string data structures are returned by the service calls. While the current service employs an experimental Glamorgan SKOS based version of DDC as the vocabulary, the service is applicable to any SKOS represented vocabulary.
The web service call interface is documented here
This includes an API tester allowing client developers to experiment with some of the calls.
Presentations and demonstrations were given at the NKOS Workshop at ECDL 2009 (ppt) and a JISC Personalisation Meeting at MIMAS, October 2009. Some aspects of the work were presented at an International UDC Seminar in The Hague, 2009 and a subsequent paper was published in Knowledge Organization 37(4) 2010.
An abridged version of the final report is available.
The project team
Andy Priest, Technical Manager, Mimas, The University of Manchester
Janine Rigby, Project Manager, Mimas, The University of Manchester
Caroline Williams, Deputy Director, Mimas The University of Manchester
Ceri Binding, Hypermedia Research Unit, The University of Glamorgan
Douglas Tudhope, Hypermedia Research Unit, The University of Glamorgan