Analysis of Evolutionary Trends in Astronomical Literature using a Knowledge-Discovery System : T etralogie
Josiane Mothe, Paul Sabatier, D. Egret, Taouq Dkaki, Bernard Dousset · 1998
Databases of electronic abstracts provide a wealth of information for studies about covered topics, collaborative works, or short-term evolution trends. In this paper, we report a study of a subset of the ADS database (Astrophysics Data System) of astronomy abstracts, restricted to 6190 articles for which at least one of the authors is affiliated to French institution, in the range 1987 to 1996. In a next step, Tetralogie is used for analysing evolutionary trends. For that, some data analysis mining functions, such as the Principal Component Analysis or classifications methods, are used to discover the most significant combinations of keywords, clusters of similar papers or outliers presenting extreme characteristics in the analysis. In addition to allowing us to discover global dependencies between the document contents, the applied methods allow one to discover features on the short-term evolution of topics along the years covered by the dataset: what are the growing fields, the emer...