Topic-centric Recommender Systems for Bibliographic Datasets

Abstract

In this paper, we introduce a novel and efficient approach for Recommender Systems in the academic world. With the world of academia growing at a tremendous rate, we have an enormous number of researchers working on hosts of research topics. Providing personalized recommendations to a researcher that could assist him in expanding his research base is an important and challenging task. We present a unique approach that exploits the latent author-topic and author-author relationships inherently present in the bibliographic datasets. The objective of our approach is to provide a set of latent yet relevant authors and topics to a researcher based on his research interests. The recommended researchers and topics are ranked on the basis of authoritative scores devised in our algorithms. We test our algorithms on the DBLP dataset and experimentally show that our recommender systems are fairly effective, fast and highly scalable.