Content analysis of documents using neural networks: A study of Antarctic science research articles published in international journals
ADVANCES IN POLAR SCIENCE · 2012
Content analysis of scientific papers emanating from Antarctic science research during the 25 years period (1980—2004) has been carried out using neural network based algorithm–CATPAC. A total of 10 942 research articles published in ScienceCitation Indexed (SCI) journals were used for the study. Normalized co-word matrix from 35 most-used significant words was usedto study the semantic association between the words. Structural Equivalence blocks were constructed from these 35 most-usedwords. Four-block model solution was found to be optimum. The density table was dichotomized using the mean density of thetable to derive the binary matrix, which was used to construct the network map. Network maps represent the thematic character ofthe blocks. The blocks showed preferred connection in establishing semantic relationship with the blocks, characterizing thematiccomposition of Antarctic science research. The analysis has provided an analytical framework for carrying out studies on the contentof scientific articles. The paper has shown the utility of co-word analysis in highlighting the important areas of research inAntarctic science.