Waseda Institute for Advanced Study (WIAS)Waseda University

News

International Political Research through Quantitative Text Analysis Kohei Watanabe, Assistant Professor

What is Quantitative Text Analysis?

Quantitative text analysis is a technique that uses computers to process written or spoken words, or “natural language,” to extract information for the social sciences and humanities. In political communication research, which is my field of study, quantitative text analysis is used for many applications, including inferring the political ideologies of candidates through analyzing election manifestos and politicians’ speeches, or uncovering concealed political bias in newspaper articles. Quantitative text analysis is also used for tasks such as classifying books according to themes and, in a different kind of application, searching for the authors of anonymously published literary works.

Research on natural language processing began with research into automated translation as an application of decryption technology during the Second World War. In the 1990s, the technology advanced dramatically due to improvements in computer performance in conjunction with the increase in the availability of electronic data and emergence of the Internet; thus, statistical analysis techniques for large-scale text data were established. From the early 2000s to the present day, a variety of machine learning models that can be used for quantitative text analysis have been developed.

Development of the Quantitative Text Analysis Tool Quanteda

I obtained a Ph.D. from the London School of Economics and Political Science (LSE), and subsequently worked there as a Research Officer before returning to Japan in 2018. During that period, I was and currently remain involved in the development of Quanteda (an abbreviation of Quantitative Analysis of Textual Data), a text analysis package written with the programming language R.
Quanteda is a tool for quantitative text analysis that Kenneth Benoit from the LSE began developing in 2012 with support from the European Research Council, and Version 1.0 was announced in January 2018. A variety of software has been developed for text analysis to date, but no multifunctional and efficient software has yet been developed.
Even at its present stage, Quanteda is more multifunctional than tm and tidytext, the previous R packages, and gensim, a Python package. It is also superior in terms of processing speed and memory usage. It is easy for researchers in humanities to use and has gained the support of leading political scientists in North America and Europe.
It is further differentiated by its ability to support Asian languages such as Chinese, Japanese, and Korean due to its compatibility with Unicode, which can handle all languages. Using the same analytic tools and models as English-speaking researchers, it is possible to make international comparisons and publish in leading journals. I think that researchers in Asia, at least those in Japan, have great expectations of Quanteda.

The Quanteda development team. Professor Kenneth Benoit is to the right of the lecturer Kohei Watanabe in the center.

Analyzing Changes in Other Countries’ Perceptions of the United States from Newspaper Articles

Using Quanteda, I analyzed how other countries’ perceptions of the United States have correlated with the nation’s domestic political situation over the past 30 years through newspaper articles published in Japan and Britain.
I performed a keyword search on a database of Japanese and British articles (using the terms “United States” OR “US” AND “government” OR “politics” OR “diplomacy” OR “military”)and used a semi-supervised vector space model (Latent Semantic Analysis)to quantitatively analyze how the 141,746 articles extracted from The Asahi Shimbun and the 176,399 articles extracted from The Guardian covered the United States from the perspective of framing.

Framing of the United States by The Asahi Shimbun

 

Framing of the United States by The Guardian

When the measured values are averaged for each term of presidents Reagan (1981 – 1988), Bush (H.) (1989 – 1993), Clinton (1993 – 2001), Bush (W.) (2001 – 2009), Obama (2009 – 2017), and Trump (2017 – ), it is clear that perceptions of the United States have changed over a 30-year period.

Changes in Perception of the United States in Japan

 

Changes in Perception of the United States in the United Kingdom

Analyzing Russia’s International Internet Propaganda Strategy

Another example of research using Quanteda is its application in analyzing articles from the state-controlled Russian website Sputnik News (https://sputniknews.com/). Sputnik News has editorial departments in more than 30 languages, including Japanese, English, Spanish, French, and German.
I classified the 51,651 English-language articles that have appeared on the website since July 2017 into six topics, economy, politics, society, diplomacy, military, and nature, and extracted the following keywords from each topic.

Topic Seed words
economy market*, money, bank*, stock*, bond*, industry, company, shop*
politics parliament*, congress*, party leader*, party member*, voter*, lawmaker*, politician*
society police, prison*, school*, hospital*
diplomacy ambassador*, diplomat*, embassy, treaty
military military, soldier*, air force, marine, navy, army
nature water, wind, sand, forest, mountain, desert, animal, human

 

The graph below shows a classification of featured article categories by country.

There are many articles on “society” and “politics” for the United States (us) and Britain (gb) because Sputnik News targets these two countries for propaganda. “nature” is a somewhat bigger topic for Japan (jp) because there are many articles related to the environment, such as those covering nuclear power. Naturally, “military” is a prominent topic in the coverage of Syria (sy) and Afghanistan (af), but the larger “society” topic for Sweden (se) gives a sense of Russia’s intention to prevent Sweden from joining the North Atlantic Treaty Organization (NATO). I presented this research at the ECPR General Conference held in Hamburg, Germany, in August.

Share of Countries and Topics in the English-language Edition of Sputnik News

Media and the Future of Quantitative Text Analysis

Media research is my original field of study. Our information environment is set to shift steadily in an undesirable direction if the media is left unchecked—a process that is exemplified by the proliferation of so-called fake news and the increasing partisanship of newspapers. In order to halt this shift, I believe that more researchers must conduct analyses from diverse perspectives and identify problems.
In this respect, I am aware of the significance of developing Quanteda as “a tool that individuals can use” to efficiently process large volumes of data that continue to be created every day without using large-scale computers or huge amounts of manpower.

Interview and Composition:Ayako Yamamoto
In cooperation with: Waseda University Graduate School of Political Science J-School

 

Page Top
WASEDA University

Sorry!
The Waseda University official website
<<https://www.waseda.jp/inst/wias/en/>> doesn't support your system.

Please update to the newest version of your browser and try again.

Continue

Suporrted Browser

Close