Exploiting Coauthorship to Infer Topicality in a Digital Library of Computer Science Technical Reports

We propose a method of mapping the topical content of distributed digital libraries and demonstrate the technique using data from the Networked Computer Science Technical Report Library (NCSTRL) digital library project. This method seeks to exploit information derived from document coauthorship to produce improved automatic subject classifications of the documents. In a distributed digital library, these subject classifications are useful in characterizing both intra-site and inter-site content. They are also helpful in providing secondary retrieval services. We present the method and describe an experiment and results showing that improved clusterings can be achieved relative to traditional document clustering. 1 Introduction Digital library projects[6] place heavy demands on existing infrastructure and hence, the performance of these systems is an object of considerable study. With the increased availability of digital libraries and distributed information retrieval systems, we can..