A Computational Analysis of Korean Women Leaders’ Narratives Using Topic Modeling

Abstract: Despite South Korea's remarkable economic growth, becoming one of the developed countries as in Organisation for Economic Co-operation and Development (OECD) in 1996, women’s status continues to lag behind. According to the World Economic Forum’s Global Gender Gap Report (2023), which shows a country’s state of global gender diversity and equality, Korea ranks 105th out of 146 countries surveyed. Korea’s gender gap index (0.68) falls below the global gender gap index (0.684), particularly in terms of economic participation (0.597), suggesting that Korea’s low rank largely comes from the lack of women leaders in government, politics, and business.
In order to address Korean women leaders’ lower status by global standards, we have conducted eight qualitative studies on the challenges they face in the workplace in their own words over the past decade (Cho et al., 2015; Cho et al., 2016; Cho et al., 2017; Cho et al., 2019; Cho et al., 2020; Cho et al., 2021; Cho et al., 2022; Han et al., 2018). From the eight qualitative studies, we collected 200 women leaders’ narratives from diverse sectors including the corporate, academic medicine, entrepreneurship, multinational corporations, higher education, and more. As a result, we have identified women leaders’ challenges in work-life balance, leadership development, and career development and success, and that those challenges are primarily attributed to the gendered workplace where organizational constraints and traditional values coexist. While our studies shed light on the challenges women leaders face in their career and leadership development, we did not examine any associations and relationships between research constructs reviewed due to the qualitative nature of the eight studies.
Although generalizability is not a goal of qualitative research, examining research topics that will be identified through topic modeling, a computational analysis method which compares research themes identified from the eight qualitative studies would give us the opportunity to revisit women leaders’ narratives from the perspective of quantitative analysis. Research (Baumer et al., 2016; Nelson, 2020) indicates that a computational analysis of qualitative data offers reproducible and efficient analytical tools by integrating a quantitative approach. Recent developments in computational analysis in qualitative research (e.g., Baumer et al., 2016) suggests a shift towards a computer-led coding process using machine learning, such as topic modeling as a commonly used method. Nelson (2020) refers to a qualitative data analysis process where a computer detects words and patterns in texts as newly coined computational grounded theory.
Computational analysis ensures a high degree of reliability, trustworthiness, reproducibility, and efficiency in the data analysis process (Nelson, 2020). In this study, we will use topic modeling as the method for computational analysis to reanalyze Korean women leaders’ narratives and compare key topics identified from this computational analysis with qualitative themes identified from eight qualitative studies. To that end, the current study is guided by the research question: how do we compare research themes identified in the previous qualitative studies with the key topics identified from the computational analysis using topic modeling?
In this study, we use structural topic modeling (STM) that charts a systematic pathway from the initial data intake to its eventual visualization, ensuring that researchers can seamlessly navigate through the algorithm-identified themes (Roberts et al., 2019). For analytical purposes, 200 women leaders’ narratives (transcripts) were concerted to a structured dataset that is compatible with the R software. We chose sentences as the basic unit of analysis, as our research focus is to unearth overarching themes and discern patterns. From the 200 narratives (transcripts), we extracted 70,529 sentences, and for further analysis, we ended up with a corpus that has 69,102 sentences and 39,454 terms.
In analysis of 200 narratives using topic modeling to compare the themes and topics identified from the previous qualitative studies and computational analysis, this study will provide implications for HRD research and practice: first, this study will provide the possibility of exploring an uncharted area for a new methodological approach to qualitative research using computational analysis, supplementing the subjectivity of qualitative research. Second, in a computational analysis of 200 women leaders’ narratives, this study will capture certain aspects of their experiences that might have been bypassed in the qualitative studies. Third, the challenges faced by women leaders across various organizational contexts, as identified in this study, will contribute to the development of policies and programs that support women in the leadership pipeline and gender equality in organizations, which has to do with the UN’s Sustainable Development Goal 5 (Gender Equality) for all women and girls (United Nations, 2015).