Our Methods

Our work builds on computer-assisted methods of data analysis. We draw on media databases such as Lexis-Nexis, ProQuest, and Factiva to analyze tens or hundreds of thousands of articles about groups such as Muslims, Jews, Catholics, Hindus, Latinos, refugees and asylum seekers, immigrants, and other status minorities. We have developed and validated lexical sentiment analysis procedures to estimate the tone of each individual article. Moreover, we calibrate the tone of articles to that of a representative cross-section of the print media; this allows us to assess the tone of articles about particular groups in a comparable way. Our method is uniquely suited to provide information about how different groups are covered.

In addition, we use collocation analysis to identify words most commonly found in close proximity to the name of the group within our sets of articles. For example, we can compare words within sentences containing the word “Muslim” to words in sentences that do not have the word “Muslims.” This allows us to see whether media outlets are especially likely to associate any group (such as Muslims) with particular words in ways that are likely to stick in readers’ minds. We also use topic modeling to discern clusters of words that are found in close proximity to one another across sets of articles. These can signal the presence of distinct topics embedded within a corpus of articles.

Using collocations and topic modeling, we can compare how different media outlets within a country or across countries discuss particular groups. Are some newspapers more prone to cover Latinos using positive words? Does the press in some countries use more negative words to discuss Muslims? What are the most common themes found in “refugee” articles, and do the differ according to whether the stories are found in the American or British press? Our methods allow us to answer questions like these.

For more detailed information about our methods and for examples of how they are applied in scholarly articles, please see our publications or request our papers in progress, and email us with questions or comments.

PAPERS IN PROGESS:

“How Negative is the Negative Coverage of Muslims? A New Approach to Sentiment Analysis for the Social Sciences,” A. Maurits van der Veen and Erik Bleich

“The Good, the Bad, and the Ugly: A Corpus Linguistics Analysis of US Newspaper Coverage of Latinx, 1996-2016,” Erik Bleich, James P. Callison, Georgia Grace Edwards, Mia Fichman, Erin Hoynes, Razan Jabari, and A. Maurits van der Veen