The Cold War and Word Embeddings: Exploring Gender in the FRUS Series, 1945-1968

The field of US foreign relations has embraced a series of methodological shifts over time. In analyzing the motivations and developments in foreign policy, historians have highlighted a variety of actors and experiences. In recent years, the field of digital humanities has entered the halls of history departments worldwide, bringing with it new tools and techniques. It is because of this development that historians of foreign policy have begun to ask new questions and re-examine the field that has adapted countless times to shifts in the humanities. By using digital methods, scholars may be able to reexamine traditional sources and test existing theories in order to conduct more nuanced analyses. Equipped with new questions and methodologies, historians of foreign relations can re-discover the collections that live in government archives or in digital collections.

As a product of these recent developments, this project uses word embedding models to investigate the contents of the Foreign Relations of the United States (FRUS) relating to the early Cold War in Latin America. The series is comprised of documents relating to various levels of policymaking, from State Department memorandums to official policy. The corpus presents an interesting case in which to explore gender. A collection that is an accumulation of major foreign relations documents, the corpus is not traditionally known for its depictions of gender. Yet gender is an inherent aspect of social relations and therefore should be present in a distant reading of the documents.

This project acts as a case study that demonstrates how humanists can use word vector models to reexamine their fields of research. “The Cold War and Word Embeddings” explores the language used by State Department officials during the Cold War. Latin American nations acted as arenas for the deployment of US Cold War policy on the local to national level. But the relationship between those writing policy and the society they were representing has not often been examined and there is still a lot to be discovered by using digital methods to understand how policy makers understood the social and cultural constructions influencing their policy. By isolating the documents concerning Latin America during the early postwar period, this project seeks to examine how the State Department understood gender within the Western hemisphere between 1945-1968.

While this project originally intended to examine the usages of gendered terms in the corporus, further investigation and acknowledgement of the limitations of such a research question encouraged a certain framework for this research. Word embeddings are powerful tools that can help scholars explore areas that are often neglected or those that need a fresh look, but they are also products of the limitations of their sources. After recognizing the construction of the textual data that this project relies on, I questioned the validity of the research questions and the texts that this project is based upon. Therefore, I use the theoretical framework of “data feminism” to situate my findings. This project demonstrates the power of word vector models, but also the constructions confronting the corpora and the researchers themselves through the lens of data feminism. These analyses are only as powerful as the contextualization of their data. Overall, this study aims to demonstrate that this corpus, and foreign policy in general, is based on a series of constructions.

To read more on word vector models and the other methods used in this project, see this page on methods and sources, and this page for the code used in my analysis.

Project Origins

This project was conceived during a workshop in July 2019. The workshop, Word Vectors Institute, was presented by the Women Writers Project at Northeastern University as an introduction for digital humanists interested in text analysis. I had previously utilized the FRUS series for a text analysis project and used the corpus I had on-hand to explore during the institute. Queries that explore class and gender are easy entries into the text and presented interesting results from the earlier generation of this project. Based on this exploratory workshop, I chose to refine the research questions and the scope of the project for further analysis.

Please also note that this project is an experimentation in nonlinear presentations. Each subpage of this project is linked to one another so users can navigate at their own pace and based on their own interests. This is intended to maximize audience presence and accessibility.