토픽모델링을 활용한 쟁점 분석

Abstract

This study aims to explore the possibility of using topic modeling in terms of issue analysis in tracking and detecting social issues. Among the topic modeling methods, LDA(Latent Dirichlet Allocation) based topic analysis and STM(Structural Topic Modeling) need to be used separately according to the research approach. STM is suitable for Confirmatory Data Analysis for hypothesis test, and LDA is suitable for Exploratory Data Analysis, such as issue analysis on news media. Dynamic data analysis may be possible if document categories are classified using meta-information in documents in LDA based topic modeling. In order to empirically examine, an LDA based topic analysis was conducted on hatred articles reported by 10 major national daily newspapers in Korea. From the results, all newspapers mainly dealt with hatred crimes against women in 2021, and hatred politics of politicians showed an upward trend in 2022. The direction of news reporting on the hatred issue did not show much difference among newspapers, and the characteristics of news reporting due to political bias were not large. This indicates that the issue of hatred has not yet emerged as an important issue in news coverage. In conclusion, this study demonstrates through empirical analysis that LDA based topic modeling can be fully utilized for research activities that track and detect issues.