Computer science research collaborator recommendation based on temporal relation and publication similarity

Abstract

A collaborator selection is an important process of entering a new quality research. Unfortunately, research collaborator recommendation is still an open research topic which has not yet been sufficiently studied. In this dissertation, a new methodology was proposed to recommend the most suitable collaborators for an inquired researcher. There are two main models which are the structure approach and the semantic approach of analysis. First, a quantitative measure called Structure Cohesion based on Collaboration Over Time (SC-CoT) model is introduced. It concentrates on the quantity of publication underlining the structure of co-authorship network over time. Second, Semantic Similarity with Background and Trend of Research (SS-BaT) model based on semantic approach has been proposed to identify a semantic similarity between researcher pairs. The content of published papers are used to assign the research knowledge in various topics for individuals before determining the research background similarity between researcher pairs. Also, a current research trend of each researcher is automatically detected. For reducing the search time, the potential collaborators are retrieved within six degrees of separation. With six years of published papers over six topic areas, the results of proposed methodology is outperformed the results of existing methods. Both SC-CoT and SS-BaT models must be comprehensively combined for recommending collaborators to the inquired researchers. The results from the various experiments convince that the methodology can recommend the suitable collaborators for a given researcher in a particular topic domain.