Understanding Knowledge Transfer from Academia to Social Media through Concept Level Analysis

Advances in social network analysis, natural language processing tools, and the availability of large online datasets have led to the proliferation of social science research on information dynamics and knowledge transfer. However, there has been little research on the cross-domain transfer of knowledge from scientific research venues to the public masses. Existing studies have analyzed the adoption of research in policy through citations and citations in new patents, but few have analyzed mass media or social media because citations are sparse in these venues. The present paper contributes the novel direction of understanding knowledge transfer at the concept level instead of the document level using computational phrase mining techniques. Specifically, I analyze the transfer of COVID-19 research concepts by correlating linguistic and social network structure features to the popularity of a given research concept. Using AutoPhrase, a text segmentation algorithm, more than 120,000 concepts were derived, and of a small sample of concepts, 67.5% were found to be transferred to Twitter. Furthermore, I propose several solutions to the current limitations of this study for ongoing and future work.