Co-Ranking Authors in Heterogeneous News Networks

Abstract

Expert Finding has been a widely studied area of research. However, most of the work in this area has focused solely on analyzing networks representing people in academia. In this work, we will present an approach for two types of heterogeneous news sources (i.e., Traditional Network Sources (TNS) and Policy Network Sources (PNS)) for experts on a set of topics. Our overall objective is to discover who are the expert journalists and policy analysts on specific topics. This work is based on our intuition that the PNS and TNS could complement each other, thus leveraging information for the learning task. We propose a probabilistic generative model named Context-based Latent Dirichlet Allocation (CBLDA) that performs the task of co-ranking authors in the heterogeneous networks of TNS and PNS. We will demonstrate that our proposed approach outperforms baselines in terms of precision, mean average precision, and discounted cumulative gain.