Last updated: Dec. 20, 2024
Overarching Research Interests
My scholarly interests focus on the dynamics of global and digital media, exploring how they shape societal discourses, reinforce norms and either perpetuate or challenge inequalities. I am particularly intrigued by strategic communication’s role in fostering positive international relations and empowering marginalized groups for social change, in addition to understand structural contradiction embedded in global politics and surrounding global communication.
Evolving Climate News Frames and Sentiment in U.S. Media:
Applying Machine Learning
My research focuses on two key perspectives on environmentalism: 1) transformative environmentalism and 2) adaptive environmentalism. Transformative environmentalism advocates for radical changes to reduce our dependence on fossil fuels, promoting solutions such as stricter regulations to maintain Earth's carrying capacity and community-driven approaches. Adaptive environmentalism suggests adapting to environmental changes through market-centered and technology-driven solutions, or internalization of these solutions into the international governance, without drastically altering our reliance on fossil fuels.
My dissertation and recent study found that adaptive environmentalism has become more prominent, while transformative environmentalism becomes marginalized. While adaptive solutions are promising and important, I believe that we need to diversify solutions, bringing more transformative perspectives into public conversation, in order to truly address climate crisis.
Market-based solutions tend to rely on economic profitability, often paying less attention to ideas or agents that generate lower financial returns. For instance, media tend to cover carbon capture technology more than community recycling initiatives or local events. Technological and financial investments are more directed to China and India than toward countries in Africa. In addition, low-income countries and communities face limited access to the necessary technologies or capital to benefit from market-based solutions.
In a similar vein, technology-driven solution alone cannot address the deeper issues of climate change. For instance, while renewable energy technologies already exist, they have not been widely adopted due to economic barriers. In many cases, they merely become additional energy sources that contribute to economic growth without displacing fossil fuels. When discussing technology-driven solutions, it is essential to consider the rate of fossil fuels being replaced by renewable energy.
Thus, I believe that diversification of perspectives (or paradigms) in addressing the climate crisis would help us achieve true sustainability by encouraging deeper, structural changes.
In this line, my research, "Frame-sentiment dynamics and evolutions in U.S. climate news: Semi-supervised machine learning and panel data analysis," analyzed the interplay between media framing and news sentiment. This study analyzed 56,475 news articles from prominent U.S. newspapers, utilizing advanced "guided" LDA topic modeling to identify four primary news frames: Market Liberals, Institutionalists, Bio-environmentalists, and Social Greens.
This methodology, an enhancement from my dissertation work, inductively identified text patterns, determined 40 keywords per frame based on the visual distance among topics, recursively implemented them in probability calculation, and then retrieved and labeled the data with the frames. Using a panel data analysis, my work revealed a notable rise in the "Institutionalists" and "Bio-environmentalists" frames with significant association with positive sentiment. In contrast, the "Social Greens" frame's prevalence decreased and was associated with negative sentiment. Findings imply the predominance of market-driven solutions in the media and declining narratives seeking climate justice and global equity.
This research was presented at the Communicating Science, Health, Environment, and Risk (ComSHER) division for the 2024 annual conference of Association for Education in Journalism and Mass Communication (AEJMC).
Currently, it is being revised for journal publication. Based on recent reviewers' feedback, I am focusing on three key methodological improvements.
First, I refined the dataset by removing news articles with fewer than 30 tokens and updating stop-words to eliminate irrelevant data.
Second, I enhanced the process of deriving seed words for guided LDA topic modeling by incorporating K-means clustering. Specifically, I employed UMAP, a dimensionality reduction technique, to map 18 topics into 4 clusters. Then I used K-means clustering to extract 120 keywords for each topic group. The Python code below displays this process of extracting seed words.
Third, instead of the traditional sentiment analysis that classifies emotions as either positive or negative, I implemented the RoBERTa model from the Hugging Face package to identify six distinct emotions: joy, fear, anger, sadness, disgust, and surprise. The model also includes a neutral emotion category.
To handle the memory-intensive tasks, I created a Virtual Machine instance on Google Cloud Platform. The specifications of the VM can be seen in the image below, and I conducted the analysis using Jupyter Notebook on the VM, with the Python code available upon request.
Cumulative Effect of Public Diplomacy Investment:
Employing Time-Series Analysis
I have explored time-series analysis. While public relations literature notes the long-term effects of public relations, few studies consider time a variable. Additionally, statistical analysis of long-term effects might require consideration of cumulative or lagged effects, meaning the influence of past interventions on future outcomes. To address these issues, I employed the Granger Causality test. In brief, this test is based on particular statistical characteristics. When two or more series are integrated in a vector model (meeting statistical assumption and based on the optimal lag length), and when co-integration (statistically verified co-movement between series data) is found, the error correction term of the vector model converges to zero. Researchers can then test the null hypothesis that the coefficient of the error correction term is zero. If the null is rejected, it indicates that the past term (coefficient) significantly explains the future term (dependent variable).
"A Time-Series Analysis of Public Diplomacy Expenditure and News Sentiment: A Case Study of the U.S.-Japan Relationship" employed PDF parsing to collect semi-annual FARA reports for 23 years, API for economic data over time, web scraping to collect news articles about Japan for 23 years from three major newspapers in the U.S., and Joint Sentiment/Topic Modeling (JST) for sentiment analysis of headlines and lead paragraphs. Then, employing the Autoregressive Distributed Lag (ARDL) model with the Bounds test approach, I discovered that the public diplomacy expenditure of Japan in the U.S. over time positively and significantly explained the news sentiment of U.S. news coverage about Japan, after controlling for the economic significance of Japan to the United States, that is, trade volume between countries over time and GDP of Japan over time.
Originating from my master's degree thesis, my previous publication, "A Time-Series Analysis of International Public Relations Expenditure and Economic Outcome," tested a relationship between public diplomacy expenditure and economic outcomes. Mainly the Vector Error Correction Model (VECM) was employed. Four countries, Japan, Belgium, Colombia, and the Philippines were selected to represent high/low expenditures in the United States and high/low GDP per capita. Mainly employing Vector Error Correction Model (VECM)—one of modeling for Granger causality tests—, I showed that the preceding expenditures of Japan, Belgium, and the Philippines in the United States significantly explained the volumes of their export to the U.S. and FDI from the U.S.
Below is the link to my master's thesis.
Doctoral Dissertation and Relevant Works
My dissertation, "The Political Economy of Media Framing in Korea: An Analysis of Korean News Coverage of Climate Change, 1995-2015," examined the shift in Korean climate change discourse post the Low Carbon and Green Growth policy, underscoring its neoliberal transformation. The study employed semantic network analysis, K-means clustering, and LDA topic modeling to analyze over 21,000 news articles, revealing a paradigm shift where traditional discourses on ecological harmony are eclipsed by the neoliberal catchphrase of "Green Knowledge."
The research revealed the shifts in semantic networks from those centered on nature and ecology to those centered on state, institutions, markets, and corporations. The shift of topics—the clusters of words—showed that more market-oriented and state-led discourse and less civic-centered and public-led discourse were found after 2008. By positioning "Green Knowledge" as the central theme of climate discourse after 2008, the neoliberal framing advocates the profitability and financializability of ideas as rational and beneficial responses to climate change. This approach marginalizes previously central narratives of ecological and human-nature harmony. My dissertation demostrates how Korean media has increasingly situated climate change within neoliberal market mechanisms and institutional strategies, post-LCGG.
Below are the links to my dissertation, the digital repository of my dissertation, and my conference paper that revisited chapters of my dissertation.
Media Framing of Urban Space - The Cheong-gye-cheon and Media
My master's thesis from Korea University, "The Production of Space and Media Discourse," analyzed news coverage of Cheong-gye-cheon in the context of changes in Korea's political and economic landscape.
I examined how Cheong-gye-cheon, which became a symbol of modernity in the 1960s after being covered by roads, was previously depicted as a hotbed of crime, a neighborhood of filth and stench, and a shantytown with tightly packed shacks that was vulnerable to fire hazards. This representation of the space provided legitimacy to the new space of Cheon-gye-cheon, with its paved roads and modern buildings.
In the 1980s, Korea's economic system rapidly transitioned from manufacturing to service industries. During this period, the media persistently labeled the Choeng-gye-cheon area as a "substandard urban facility." In particular, news media frequently highlighted the traffic congestion and noise caused by trucks loading and unloading goods at the tool shops at Cheong-gye-cheon.
By 2003, environmental discourse began to emerge, initiating discussions on transforming Cheong-gye-cheon into a natural space, with the removal of the elevated highway being central to these discussions. However, this environmental discourse was "appropriated" by Korea's political-economic structure, leading to media representation that portrayed the consumption and service experiences in the "new" Cheong-gye-cheon area as an experience of enjoying "nature" and the "environment." In other words, the media framed a commercial and service-oriented space as a place offering ecological and natural experiences to the public. This process obscured the underlying commercial intentions by appealing to environmental and natural ideals.
Consequently, the "substandard urban facility" discourse, which had been continuously raised since the 1980s, lost its force with the removal of the "road" that once united the small-machinery parts sellers and the space that had symbolized the modernity of Korea. In 2005, Choeng-gye-cheon finally transformed into a consumer space lined with cafes, shopping malls, and restaurants along the waterway.
Below are the links to my thesis and published article, both written in Korean.