Data analysis of government-influenced news outlets

Context

In this project I worked with scraping and data analysis of news articles for the first time. I set out to critically compare the framing of news content between Ukrainian and Russian news outlets during the beginning of the war in Ukraine. As information wars have become central in shaping public discourse, understanding how news frames are constructed and how they differ across national lines is important. Framing becomes more prone in current society and directly impacts polarization.

I focused on two prominent, government-influenced outlets: RT News from Russia and UkrInform from Ukraine. Both play roles in communicating narratives about the conflict, yet each is embedded in a unique political and cultural environment. My aim was to analyze and visualize the subtle, and sometimes not so subtle differences in how these outlets present the news, informing about bias and the shaping of public perception.

Research & Insights

The project began with a theoretical framework grounded in media studies and political communication, particularly the concept of framing as described by Entman (1993) and Valkenburg (2000). I examined how media outlets “define problems, diagnose causes, make moral judgments, and suggest remedies” all inevitably influenced by political context and editorial stance.

To compare RT News and UkrInform, I gathered a large dataset: 2089 articles from RT News, covering March to April 2023 (the period of Russia's spring offensive), and 9598 articles from UkrInform, covering a broader period from March 2022 to April 2023. These outlets were chosen because they are both subject to direct or indirect government control, making their coverage especially relevant for a study on agenda setting and bias. This difference in sample size and time period does impair the validity, but it was fun to explore different data gathering methods.

The main research question was: How do the generic and issue-specific frames in news content differ between Ukrainian and Russian news outlets? I operationalized “generic frames” as commonly used journalistic perspectives: Conflict, Human Interest, Responsibility, Economic, and Morality. “Issue-specific frames” were identified through key nouns in the articles, highlighting recurring topics, roles, and entities.

Design Solution

The analysis relied on a data-driven approach. First, I scraped the full corpus of articles from both websites, overcoming access limitations (especially for RT News because it is banned in European countries) with a VPN, Python's BeautifulSoup and Selenium libraries. The raw texts were cleaned and preprocessed with NLTK, involving stopword removal, stemming, and lemmatization.

Generic frame classification was performed using a dictionary-based approach. Articles were categorized based on the presence of frame-related keywords. Mentions like “war”, “military” and “guns” for instance, indicated if the 'Conflict' frame is used in the article. For issue-specific frames, part-of-speech tagging extracted and counted nouns to reveal dominant themes in the coverage.

To make the results accessible and engaging, I designed an interactive Streamlit dashboard. This dashboard visually compared the frequency of frames across outlets using bar charts and enabled users to search for specific terms or frames, encouraging exploration and transparency. You can try it out below by interacting with the dataset (select fullscreen to navigate with more ease).

The results were interesting:
• Both RT News and UkrInform relied heavily on the “Conflict” frame, but UkrInform did so even more (72% vs. 50%).
• RT News emphasized “economic” and “morality” frames more than UkrInform, while UkrInform's coverage was more focused on direct conflict and local events.
• Issue-specific frame analysis showed RT News frequently referenced traditional, governmental entities like “Russia,” “Country,” “President,” and “State,” reflecting a top-down, authoritarian perspective. UkrInform, meanwhile, focused more on “Ukraine,” “report,” “attack,” “defense,” and “people,” reflecting local events and civilian experiences.

Reflection

This project offered me insight into the ways news framing is shaped by political context and editorial goals, especially in times of conflict. By combining text analysis with visualization, I was able to move beyond anecdotal observations and present some insight in bias and agenda-setting on both sides. The process showed the importance of data transparency and critical media literacy, especially as audiences are increasingly exposed to competing narratives online.

I especially liked the different data science techniques I had to use to get some results and visualize these in comprehendable insights. I do have to add that this case study cannot be seen as empherical because the sample size of the articles and the time span of the collected data is not comparable. RT News only offered 2089 articles between 2023 march and 2023 whilst the UkrInform platform offered 9598 articles between march 2022 and april 2023. In the end it was more of an experiment what I could achieve with scraping and data analysis.