Text Big Data Study on Extracting Topics from the Employee Reviews: Focusing on Exploring the Association with Job Turnover

Abstract

This study aimed to extract topics from text big data on the site Jobplanet, and explore the associations between the topics and Job Turnover. Using Latent Semantic Analysis (LSA), 35,031 employee reviews from the top 50 market capitalization organizations were analyzed. After applying a Varimax rotation, 62 topics were finally interpreted and individual topics were grouped into 8 topic groups based on the similarity of their meanings: 'Autonomy for Hours of Rest', 'Growth', 'Comparison', 'Organizational System', 'Organizational Climate', 'Pay Satisfaction and Work Intensity', 'Perceived Organizational Support', and 'Job Characteristics'. As a result of exploring the relationship between each topic and job turnover through Generalized Additive Model (GAM), there were 31 topics showing a linear relationship with turnover, and with 14 topics showing a nonlinear relationship. This study discovered 'Autonomy for Hours of Rest' as a concept distinct from other constructs studied in the prior job turnover research. In conclusion, we discussed these findings, limitations, and implications for future research.