Data mining and exploration of the Nuclear Science References

Abstract

The Nuclear Science References (NSR) is a carefully curated bibliographic dataset focused on nuclear science literature. A domain-specific search engine and supporting tools have been developed to aid and encourage the exploration of the NSR. User queries are analyzed to form a series of filters to retrieve relevant NSR entries from a database. The resulting information is presented in multiple views including lists, bar charts, and network graphs. The network graph representations offer unique insights on collaborations centered around a given parameter such as a nuclide or group of authors. The capability of clustering algorithms to expose trends within the dataset is demonstrated by clustering authors based on publication traits. A vector space model based on the metadata provided in the NSR is used to recommend semantically similar NSR entries. The completed work serves as both an example and a framework for future analysis of the NSR.