Abstract

Studies that identify research trends and capture characteristics within a specific field play a crucial role in forecasting the field’s future and guiding researchers’ directions. Recent trend-analysis studies have relied on title-and-abstract text or metadata analysis, focusing on broad topics or on key researchers and institutions. In this study, we propose a methodology that employs automatic summaries of full texts-instead of titles, abstracts, or metadata-to examine substantive aspects such as methodology, detailed subtopics, results, and discussion points. For papers collected on the topic of the digital divide, we generated automatic summaries by structuring the full text into Introduction, Methodology, Results, and Discussion using ChatGPT and then compared outcomes through word-frequency analysis, co-occurrence analysis, and keyword network analysis. The findings confirm that analyses of title-and-abstract texts and Introduction summaries reveal similar major themes. In contrast, analyses of the Methodology, Results, and Discussion sections uncover detailed information that was previously difficult to identify.