Mining and Characterizing Scientific Communities in DBLP

around specific conferences based on either common topics or common authors. Our findings show that the authors provide a better ground for characterizing communities than topics alone. At the second level we
decompose the computer science area into 14 research fields based on previous researches and study various properties of the corresponding communities. We characterize each research community in terms of
publication growth rate, collaboration trends and population stability. All these features influence each other to some extent and together show the difference in activity rate andinterdisciplinarity of each subarea.