<p>Shifts in media niches have transformed the landscape of scientific communication, with public participation on social media introducing new dynamics in the dessimination of scientific knowledge. Different user types contribute unevenly to the structure and progression of diffusion networks, influencing how scientific information circulates across audiences. This study employs a semi-supervised machine learning method based on Mix-Text to classify user identities (e.g., academic and non-academic participants) and combines social network analysis, complex network metrics, and statistical analysis to examine their roles in diffusion processes. The findings show that academic users are more influential in shaping viral diffusion structures, whereas social audience members tend to participate more actively during the later dissemination phases. Additionally, the predominance of neutral assortativity within both broadcast and viral diffusion networks suggests a high degree of cross-group interaction within the dissemination process. By clarifying how different user types shape network structures, this study contributes to a deeper understanding of the structural characteristics that underpin the online spread of scientific knowledge.</p>