User-Centered Recommender System

Work on this project has been completed.

This work was a joint research project of OCLC Research and the Information School, University of Sheffield to investigate the development of recommender systems for the retrieval of journals, books, digital media, video, etc. in, a cloud-based, multi-institution, international catalog.

Librarians often express the potential benefits of effective recommender systems based on circulation data. However, in the current economic climate it is difficult for librarians to find sufficient time and resources to build their own library catalog recommender systems. This is compounded by the users of library services wanting and expecting the same kinds of services they find in other online services, which only can be generated from large amounts of data, which typically is not easily produced and made readily available to librarians.

Librarians also are under increasing pressure to develop new services and applications that are tailored toward individuals and communities of users. Effective recommender systems may help to accomplish this. Such systems could help promote library collections by guiding users towards less widely-accessed or popular content.

Although OCLC has a recommender system for, the functionality is basic and often does not provide logical recommendations to end users. This project will enable OCLC to develop a new approach in a more theoretical and empirical manner, based on data held by OCLC. User-centered design and empirical evaluation of a prototype system would provide invaluable data for OCLC in assessing the value of recommender services for

The main objectives for this research are to:

  • Establish user needs and expectations from library-based recommender systems;
  • Perform an analysis of existing systems for library book and media recommendations;
  • Establish criteria for making and assessing library-based recommendations;
  • Analyze the data resources held by OCLC, including bibliographic records, transaction logs and library holdings data;
  • Design and build an interactive recommender prototype for the "universal" library catalogue; and
  • Evaluate the prototype in a lab-based setting and operational environment.




For the end users of, effective recommender functionality will assist with information discovery within the library catalog. The global nature of the catalog means that recommendations could be made outside of the holdings for a single institution and the amount of data held by will enable statistical-based approaches to create automated recommender functionality that is more reliable and effective. This will help meet users’ expectations of online library catalogs by providing similar functionality to other online services.

The general web community will benefit from effective recommender functionality in In services such as Google Books, browsers link to through "Find in a Library", thereby guiding web users to a universal library catalog. Recommender system functionality would highlight other items of potential interest to users, based on their user profiles and current searches and support information discovery and users' serendipitous information-seeking behavior.




The results of research carried out in this project will be made available through dissemination in academic publications and presentations at relevant conferences; through the development of a prototype recommender system based on data derived from that will be tested by a range of potential end users and librarians.


Most recent updates: Page content: 2015-08-11