Abstract
In the field of social research, there have been quantitative analyses of various kinds of textual data, such as newspapers or open-ended survey questions. In most of these cases, either the correlational approach or the dictionary-based approach has been employed. In the former approach, multivariate analyses are utilized to examine word correlations or co-occurrences and discover themes. For example, cluster analysis of words is used to find groups of words that appear frequently in the same document. In the latter approach, words or documents are assigned to preexisting categories following rules that are created by the researchers, and categorization results are analyzed quantitatively. In this paper, the author evaluates these two approaches and proposes a third approach in which they are complementarily coordinated. In order to perform analyses that follow this new approach, original software was developed and distributed as free software. The author presents an analysis of responses to open-ended questions that uses the new approach and this software. In the conclusion, the author highlights the advantages of the new approach in comparison with the traditional two approaches.
