The act of distant reading allows scholars to explore much more data than they would have been able to manually examine, but also encourages them to ask new questions or challenge the results of previous studies. In a period when humanists, and historians especially, are challenged to stay ahead of the ever-shifting field, digital methods enable scholars to do such work. As shown in this case study, the tool of word embedding models, and text analysis in general, can add to the discussion surrounding gender in foreign relations by analyzing the language used in a large corpus such as the Foreign Relation of the United States series.

In this case, these models were able to produce unexpected results and ones that provide a look at how diplomats and policymakers understood gender constructions while at their posts. The corpus concerning Latin America in the Cold War seems to suggest that the female gender is a part of an “othering” by officials. Terms that are related to civil unrest and violence between civilians and state law enforcement closely relate to terms of the female gender. A close reading based on these results shows that women and girls were subgroups of the groups at the brunt of these clashes. Other terms that closely relate to female terms or roles appear to be in close proximity to other groups affected by the instability, showing that women and girls are considered like these “other” groups within the corpus. Terms like “jail” appear in the context of unrest, without the presence of female actors nearby in the text. In the same corpus, terms of male gender almost always present a strong relationship to terms depicting roles, both professional and social. It appears as though diplomats and policymakers understood men based on their career or their social status, while viewing women as a group affected by unrest.

In the same period, officials seem to view women and girls as relating to agriculture or the family in documents relating to Europe. Familial roles closely relate to other immediate family titles, as expected. Looking at terms of male gender in the same corpus, results point to professional roles. Recognizing that this corpus is formed from the writings of individual planners and diplomats, it appears as though State Department officials recognized a very different social reality in Latin American than in Europe. Whereas men consistently are understood in terms of their roles in a professional setting or social setting, women seem to occupy different spaces between the two regions. In Latin America, officials viewed women for their place in unrest, while women in Europe were recognized for their domestic roles.

It is important to acknowledge the boundaries in which the data was collected and manipulated, and also to provide the context in which the query results can be found. It is important to provide the social and cultural constraints placed on the corpus and to recognize the series of constructs placed on the raw text. In the case of this study, the data was collected from (predominantly male) policy planning officials and executors and carefully compiled by the Office of the Historian. There are many holes in the corpus, excluding documents that were considered unimportant, but that may have had additional information on the social relations in the regions that they chronicle. With this in mind, it is equally important to provide the textual context that the vector terms appear in. For this, it is important to conduct a close reading based on the results of the distant reading. The connections might be different than scholars expect.

This project has acted as a basic gender analysis using word vector models. The future of this project will likely expand the scope of this analysis by exploring different regions and time periods. Of course, there is much more to be discovered using word vector models, and I hope this project demonstrates that to its users.

Acknowledgements

Special thanks to the Women Writers Project (WWP) at Northeastern University. This project was conceived during a workshop presented by the WWP in July 2019. Laura Johnson, who worked for the WWP, guided me through the process of cleaning my data in its earliest form, necessary for later iterations. Also, my appreciation goes to my project advisors, Dr. Julia Flanders and Dr. Sarah Connell, for their continued support since the inception of this project. Despite various obstacles, from minor code bugs to Covid-19, both Dr. Flanders and Dr. Connell quickly answered questions and provided useful suggestions at each milestone of this project.