Retrieval of Web Resources Using a Fusion of Ontology-Based and Content-Based Retrieval with the RS Vector Space Model on a Portal for Japanese Universities and Academic Institutes

This paper describes the search functionalities of JuNii Plus, a new version of JuNii, a portal site providing access to online information resources produced in the digital libraries and institutional repositories of Japanese universities or other academic institutes. Currently there are 268 member institutes. We have developed a multifaceted hierarchical controlled vocabulary for better navigation and a new search engine based on fusion of the Relevance-based Superimposition (RS) model (an extension of vector space models) and metadata-based retrieval (the ranking score is calculated based on distances on the ontology of the controlled vocabulary). By introducing content-based retrieval using the RS model, in addition to the metadata elements, texts in the metadata description and body text of the resources themselves can be used for retrieval. This is especially helpful in the retrieval of resources with poor metadata.