Comparisons of the Economist Topics on Three Countries from 1991 Through 2016

Abstract

Abstract New topic modeling technique has been increasingly used in research of communication for quick discovery of latent topics that are spread across huge volumes of text. This work intends to analyze and compare the topics automatically generated by Latent Dirichlet Allocation (LDA). The data for building LDA model in this work is based on 38,124 articles published from 1991 through 2016 in one of the world’s most influential political and economic magazines, The Economist. The retrieved documents for generating topics are divided into three countries of the UK, the US, and China in order to observe topical differences between these ingroup or outgroup countries in The Economist coverage. The work analyzes interpretability, overall weight distributions, and historical changing patterns of the topics using LDA model diagnostics. It discusses the hot or increasing trends using regression coefficient. The work also tentatively explores the relationship between the media agenda and events.