Latent topical authority indexing

Abstract

Much of scientific progress stems from previously published findings, but searching through the vast sea of scientific publications is difficult. We often rely on metrics of scholarly authority, such as the h-index, to find the prominent authors, and use these authors as starting points and waypoints. However, these authority indices do not differentiate authority based on research topics, so additional effort is needed to find the prominent authors in the relevant research topics. This thesis presents Latent Topical-Authority Indexing (LTAI), a Bayesian topic model that discovers the topical authority of scholars by jointly modeling the topics, authority, and citation network. In this thesis, four academic corpora are fitted to LTAI: CORA, Arxiv Physics, PNAS, and Citeseer. This thesis shows that explicitly modeling topical authority leads to improved accuracy over other recent models when predicting citations and authors of publications.