Emotional Analysis with News Using Text Mining for Framing Theory

Abstract

Framing theory posits that the media tend to present a constructed “reality” by selecting and highlighting particular aspects of reality while obscuring or omitting others. Hence, this process leads the audience or readers to a particular understanding of reality. Depending on how each event or issue is defined, the same event can be presented and understood in different ways or through different frames. “Traditional” or typical framing research has used content analysis to identify such frames by examining major news sources that each story adopted, words used, and the tone of the stories. However, this study aims to extend this framing research to a different level by using computer-assisted analysis. This new method allows us to analyze massive data and to visualize the representation through text-mining, natural language processing, and emotion lexicon. Within the broad framing approach, this study intends to show how major newspapers of several countries depict leaders of North Korea and South Korea. The following research questions are addressed in this study: How each leader of two Koreas is represented by different countries’ press? What image is dominant? How similar or different is the portrayed image? A total of eight newspapers written in English from six countries were selected for this case study: The Chosunilbo, The Korea Times, The Hankyoreh from South Korea, Uriminzokkiri from North Korea, The New York Times from USA, The Globe and Mail from Canada, People’s Daily from China and Thanh Nien Daily from Vietnam. Each of these countries was selected based on the history and geopolitical relations with two Koreas. Using a keyword search such as Jong-un Kim, Guen-Hye Park, national leader, or president, we identified relevant articles from these eight newspapers and analyzed them by using text mining and Natural Language Processing (NLP). Emotion analysis is a Lexicon-based method that can detect different emotions expressed in news stories. We assumed that various degrees of emotions associated with each leader would indicate the nature or orientation of reported frames. For this analysis, NRC Emotion Lexicon dictionary developed by Paul Ekman theory who identified a total of six emotions with 14,182 words was used. Then, the sentence score for each emotion was calculated. The significant contribution of this study is to present a new method of text mining and big data analysis for framing studies and to show how the overall media frames could be visualized for clear and better understanding of media representation.