In order to evaluate these propositions, the DataLab will help map the historical trajectory, prominence, and moral valence of strategy and strategists in American political life using computational text analysis. The project centers on two periods: 1860-1930, and 1960 to the present. DataLab will create a collection of texts, or corpus, of contemporary news from these time periods. Using this corpus the DataLab will work with Dr. Mudge and her research assistants Max Montrose and Lindsay Maurer to generate temporal topic and word embedding models to follow the prevalence and direction of political strategy and strategists during moments of political-economic crisis.
Topic models generate scores for each piece of text in a corpus which indicate how important the model thinks a topic is within a single text in comparison to others in the corpus. The topic models in this project will be used to measure the salience of political strategy topics and political strategists over time, as well as organizations, people, and places that are particularly central to these topics.
To track the moral valence of strategy and strategists over time, the project will also make use of word embeddings. Word embeddings are numerical representations of a word, generated by examining each word in the context of all other words in a corpus. This comparison allows words to be “mapped” as a point in a cloud with other words, with words clustered together in this cloud having similar meaning, and how those meanings change over time. The word embeddings model for this project will examine how terminologies related to political strategy changed over time, and their moral connotations. We will also follow the relationship between strategy, market, and democratic terms.