Meetings are a ubiquitous occurrence experienced daily by employees across countries and occupations, functioning as a critical procedural mechanism to enhance work efficiency and collaboration.Reflecting this practical importance, empirical studies on topics such as meeting effectiveness and meeting-induced stress have been accumulated over the past several decades in fields including organizational psychology, management, and communication.However, much of the existing research has focused on isolated time points or interpreted meeting effectiveness within limited contexts, and the onset of COVID-19 has further accelerated significant transformations in meeting practices.Consequently, there is a need for macro-level analytical research that examines how trends in meeting-related studies have evolved, identifying which topics have emerged or diminished over time, and providing theoretical and practical insights for the future growth of meeting science (Allen & Lehmann-Willenbrock, 2022).In response to this need, the present study systematically analyzed 381 articles on workplace meetings published from 1950 to the present, utilizing large language models (LLMs) to extract key nouns, construct a semantic network, and explore diachronic topic shifts.Method First, the articles were classified into five research categories-Organizational Behavior (OB), Communication, Information and Communication Technology (ICT), After Action Review (AAR), and Others-through cross-validation between ChatGPT-4 and the researchers.Next, the articles in each research category were divided into five time periods: the entire period, 1950-1999, 2000-2009, 2010-2019, and 2020-2025.In the data preprocessing phase, the natural language processing (NLP) tool spaCy was used to extract Nouns and Proper Nouns from the texts.To understand semantic relationships among major concepts and analyze the structural associations between research topics, a semantic network analysis was conducted.Co-occurrence relationships were defined, and word frequencies were calculated and assigned as edge weights within the network.For each time period and research category, the top 20 keywords were visualized as circular nodes, and their co-occurrence relationships were depicted as connecting edges.Finally, to extract meaningful topics and their associated core keywords, topic modeling was performed using the Top2Vec algorithm (Angelov, 2020).Top2Vec generates document vectors and automatically clusters them using HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) to identify topics.To observe the temporal changes in keyword appearances within each topic across the five research fields, a ranking based on keyword frequencies was created, and the top 10 keywords for each period were identified. ResultsThe visualization of the semantic network analysis for all 381 articles is presented in Figure 1.'Meeting' showed particularly strong connections with 'work', 'organization', and 'leader', indicating that a substantial proportion of studies focus on workplace contexts and highlighting the critical role of leadership in meetings.Additionally, the strong interconnections among keywords related to virtual meeting, such as "videoconference," "Zoom," and "fatigue,"
Lee et al. (Wed,) studied this question.