Employee perceptions of total rewards revealed through natural language processing of LinkedIn discourse

Abstract

This study employs an innovative methodological framework combining natural language processing (NLP) and social media analytics to examine employee perceptions of total rewards through LinkedIn discourse. Analyzing 42,852 comments collected during 2023, we implemented a multi-modal analytical approach incorporating BERT-based sentiment analysis, topic modeling, and named entity recognition. Our research addresses the critical gap in understanding real-time employee sentiment toward compensation and benefits through social media discourse analysis. The methodological framework achieved robust performance metrics (F1 = 0.83) in analyzing compensation-related discussions. Results reveal significant industry-specific variations in sentiment (F(2,42,849) = 23.45, p < .001), with technology sector discussions maintaining the highest positive sentiment (M = 0.42, SD = 0.31). Temporal analysis identified a substantial shift toward work-life balance discussions (+ 27%) and decreased emphasis on traditional compensation topics (− 12.0%). Named entity recognition successfully identified 3427 organizations, 892 benefit types, and 1245 job roles, enabling granular analysis of relationships between organizational characteristics and employee sentiment patterns. This research contributes to both theoretical understanding and practical applications in compensation research through: (1) establishing a validated framework for analyzing social media discourse in compensation contexts, (2) providing empirical evidence of evolving employee preferences in total rewards, and (3) offering actionable insights for compensation professionals in strategic benefit design and communication. The findings have significant implications for organizations adapting their total rewards strategies to align with contemporary workforce preferences.