Sažetak: A model of multi-perspective article description (MPAD) was explored and preliminarily tested. The model assumes that journal articles should be described for bibliographic purposes from three different perspectives, i.e. by using keywords extracted from: (1) article titles and abstracts, (2) titles of their cited references, and (3) titles and abstracts of articles citing them in the future. In order to explore the relationships among keyword types and to test the model preliminarily, a method labeled as Multistage Indexing of Subject Headings (MISH) based on the Keyphrase Extraction Algorithm (KEA) was employed to provide all three types of keywords for all articles from the sample. The articles were sampled from SCIndeks: The Serbian citation index. Three separate maps of (local, peripheral) science were constructed, each based on a different type of keywords. The Partitioning Around Medoid method (PAM) for cluster analyses, followed by multidimensional scaling for visual representations of extracted clusters, was employed. Results suggest that the three types of keywords generate relatively similar maps, encouraging keywords aggregation for practical purposes. Some differences among the maps are not fully consistent with the predictions derived from the model. They reveal some methodological deficiencies of the study and indicate the most promising directions for further research.

Ključne reči: citations titles; models; science; maps; indexing; articles; abstracts; references; clusters

Puni tekst (Engleski)