Volume 4 Issue 1
Spring 2008
ISSN 1937-7266

FacetedDBLP - Navigational Access for Digital Libraries

Jörg Diederich and Wolf-Tilo Balke

L3S Research Center
Leibniz Universität Hannover
Appelstr. 9a
30167 Hannover
{diederich , balke}@L3S.de

Categorization systems (also known as classification systems or taxonomies) have been found very useful, for example, for navigation in large digital libraries or for browsing very large result sets of a keyword search. Enriched by additional metadata like publishing year, or type of publication, categorizations allow users to effectively filter collections to suit their information needs.

Users of Digital Libraries can have a broad spectrum of objectives when they access a document collection. For example, some users search for a specific document, others search for relevant documents about a specific topic, and still others look for new documents in a certain area. Furthermore, some users are also only interested in a summaries of the results, and not necessarily in individual documents. This is typically the case when users want to learn about related topics, people devoted to work on a specific topic, or relevant publishing venues for a particular topic. The faceted search paradigm offers a well-founded compromise to support the diverse information needs of most of these users using a single user interface.

The L3S FacetedDBLP system provides a faceted search interface for the publication metadata available in DBLP. These facets can be used in several different ways: First, they provide summarized information on query results, for example persons or publication venues related to the query. Second, facets can be used to filter the result set e.g., for a specific year or author, or any combination of all the available facets.

A special feature of FacetedDBLP is the so-called GrowBag facet. GrowBag graphs allow query results to be represented in the form of communities. GrowBag graphs show the main topics associated with the documents in the results set and enable users to even filter the results for a specific community. In this way the GrowBag graph provides an automatically created and maintained light-weight categorization system. The use of GrowBags was experimentally evaluated in different communities like for instance over the Medline dataset in the medical domain and was found suitable especially suitable for small or highly dynamic communities. Respective evaluations can be found here.

Currently the L3S FacetedDBLP system has attracted users from all over the world with more than 5.000 hits per day (as of April, 2008) and is cross-linked from popular sites such as the original DBLP site.