collection management;
โข organizing and indexing the materials for storage
We find that dissemination of collection-wide information (CWI) in a distributed collection of documents is needed to and retrieval; achieve retrieval effectiveness comparable to that of a central-
โข user interfaces and human-computer interaction; and ized collection. Complete dissemination is unnecessary. The โข interoperability of disparate heterogeneous systems.
required dissemination level depends upon the content-skew of the distributed collection, i.e., how documents are allocated
We concentrate here on organizational aspects of a digiamong sites. Low dissemination is needed for random document tal library that bear on the retrieval performance of the allocation, but higher levels are needed when documents are system. From that perspective it is useful to consider the allocated on the basis of their content. We define parameters digital library as a distributed information retrieval (IR) to control dissemination and document allocation and present system. As in traditional centralized IR systems, the perforresults from four document collections. These results provide mance of such systems is gauged along two dimensions: insight into the necessary technology underlying digital librarefficiency-how quickly a query is processed; and effectiveies. We also describe the architecture of the Networked Comness-the quality of the resulting response. In this paper, puter Science Technical Report Library (NCSTRL), a concrete we are concerned with the latter dimension, retrieval effecexample of a system that fits our model of a distributed artiveness. How can we achieve effectiveness comparable to chive.