Data Extraction and Integration for Scholar Recommendation System

Abstract

Recommendation systems have been an integral part of massive open online courses (MOOCs). With a large amount of availability of data and resources, recommending scholars and professors through general reviews and academic advisor applications has become a tiresome job. Finding professors and scholars relevant to a student's area of interest involves a combination of multiple factors like field of study, depth of research area, research background of professors, ongoing research opportunities, etc. As recommending scholars and professors deals with so many different factors, it is very complex and unreliable when done manually. In this paper, we present a content-based mining approach to go through all relevant resources, extract required information, and use it to recommend a list of scholars based on student's area of interest. For our experimental model, we gathered information about a number of professors at our institution from various web resources such as IEEE, Springer, ACM, Sciencedirect, arxiv and department website. We use topic modeling and clustering algorithms in our content-based mining approach. We present a comparative analysis of the following topic model algorithms: latent dirichlet allocation (LDA), hierarchical dirichlet process(HDP), latent semantic analysis (LSA) and clustering techniques: k-means and hierarchical clustering in determining the most accurate recommendation list of professors or scholars.