Finding rising stars through hot topics detection

Abstract

Topic modeling methods have usually been applied in the past to identify the research interests of researchers. Observing the scientific growth, the trending topics can be identified as Stable, Hot, or Cold. Finding rising stars (junior researchers, who are at the start of their career) from a bibliometric network is a challenging task, specifically if the researchers have an interest in multiple sub-domains or are working on diverse topics. Existing methods for finding rising stars explore the co-author networks or citation networks, and ignore the textual content, which may help in finding rising stars through hot topics detection over time. A publication contributing to a hot topic can be an indication that the author of that publication may be a rising star and can become an expert in that domain in the future. This study proposes the Hot Topics Rising Star Rank (HTRS-Rank) method for finding rising stars by detecting hot topics. HTRS-Rank finds the junior scholars, who contribute to hot topics at the start of their career and ranks them based on the presence of hot topics in their publications. AMiner five years dataset ranging from 2005–2009 is selected for experimentation. Top 10 researchers are considered to measure the association strength using rank correlation among HTRS-Rank and baseline methods. Experimental results show the efficiency of HTRS-Rank in comparison to the baseline methods. The proposed HTRS Rank (TF–IDF) provides low standard deviation for productivity, citations and sociality as compared to baseline methods for more social and highly cited authors. It is identified that HTRS-Rank (WordNet) emphasizes the semantic similarity of two sentences, whereas HTRS-Rank (TF–IDF) scheme emphasizes the uniqueness or importance of each term, therefore TF–IDF approach performs better than WordNet approach due to having higher correlation with StarRank and WMIRank.