Annotation: During the last decades scientometric techniques have been used for research activity stimulation. Number of published articles and number of their citation counts are among the most important scientometric parameters. In an automated environment, when the publications metadata is gathered from various sources, correct linking of original papers with their translations into different languages is extremely important. In the paper we show that the known text similarity measures are inefficient in the context of article linkage problem. We propose a method for semi-automatic article linkage using statistical data on authors publication activities only. This approach may be used for linking articles without training for the language of translation. The method was evaluated on real-world collection of publications metadata of ISTINA information system.
Keywords: bibliographic data, graph analysis, translation, article, statistics, scientometrics, citation, automated systems, библиографические данные, анализ графов, перевод, статья, статистика, наукометрия, цитирование, автоматизированные системы