Online attention versus knowledge utilization: Exploring how linguistic features of scientific papers influence knowledge diffusion

Knowledge diffusion breeds technological innovation and promotes scientific development. In modern times, knowledge is disseminated in both the academic community and on social media. Despite a rich body of researches on factors influencing knowledge diffusion, they pay less attention to linguistic features and mechanisms behind different kinds of knowledge diffusion. To address the research gaps, this study empirically examined how linguistic features (lexical density, lexical sophistication, syntactic complexity, and cohesion) of scientific papers influenced their diffusion. Moreover, we compared the various roles of linguistic features in two knowledge diffusion mechanisms, that is, knowledge utilization and online attention. We first proposed our hypotheses based on the construction-integration model, attention theory, cognitive load theory, and social capital theory. Then, using normalized citation (NC) and normalized altmetric attention scores (NAAS) to measure knowledge utilization and online attention, respectively, regression models on full texts of research papers from PLoS were constructed to identify the effects of linguistic features on knowledge diffusion. Through delicate analyses, the inverted U-shaped relationships between lexical density/lexical sophistication and NC were identified. But cohesion/syntactic complexity had no statistically significant effect on NC. In addition, the U-shaped relationship between lexical density and NAAS, and the inverted U-shaped relationship between lexical sophistication and NAAS were verified. A positive relationship between cohesion/syntactic complexity and NAAS was also found. Then, to further discuss whether there are domain differences in the impact of linguistic features on knowledge diffusion, we conducted a heterogeneity analysis. On the theoretical front, this study provides fresh perspectives to knowledge management literature by incorporating linguistic features and to scientometrics literature by comparing the distinct effects of linguistic features on NC and NAAS. Additionally, this study extends the application of several theories including construction-integration model and cognitive load theory. On the practical front, this study offers insights into the sensible way of academic writing and knowledge promotion.