Indexing Institutional Data to Promote Library Resource Discovery
This article introduces a recommender system designed by the NCSU Libraries to increase the discovery and use of relevant academic library resources for open-ended topical searches in a variety of search contexts.
The Smart Subjects system exploits the rich topical content present in scholarly oriented institutional data stores as the basis for generating subject recommendations. The NCSU course catalog and the NCSU Faculty Publications Repository are currently being used as data sources. Content is extracted from these sources to generate text extracts representing NCSU academic departments from teaching and research points of view. These text extracts are indexed to provide keyword-based relevance-ranked retrieval within the collection. Subsequent steps involve applying scoring weights to search matches from the two data sources, and a final crosswalk between academic department codes and the homegrown library subject classification to generate the recommended list of library subjects. This approach enables us to generate subject recommendations using arbitrary text, such as user search queries, as input.
It is intended for the Smart Subjects system to eventually support a variety of lightweight library resource discovery services. The first application is a simple library subject portal discovery service that is integrated in the library website search tool. Future applications include a service to help users pre-select topically relevant metasearch targets when searching for articles in a metasearch environment. Work is underway to improve the quality of the recommendations by increasing the size and variety of topical content indexed.
© Copyright 2007 Tito Sierra