Mining Research Topic-Related Influence Between Academia and Industry

Abstract

Recently the problem of mining social influence has attracted lots of attention. Given a social network, researchers are interested in problems such as how influence, ideas, information propagate in the network. Similar problems have been proposed on co-authorship networks where the goal is to differentiate the social influences on research topic level and quantify the strength of the influence. In this work, we are interested in the problem of mining topic-specific influence between academia and industry. More specifically, given a co-authorship network, we want to identify which academia researcher is most influential to a given company on specific research topics. Given pairwise influences between researchers, we propose three models (simple additive model, weighted additive model and clustering-based additive model) to evaluate how influential a researcher is to a company. Finally, we illustrate the effectiveness of these three models on real large data set as well as on simulated data set.