History Meets Computation: A Case Study of Computational Text Analysis on 18th-Century Ottoman Court Histories

This research project began with a deceptively simple question: How did Ottoman court historians, as members of the imperial bureaucracy, reflect upon and legitimize the transformations of their Empire in the long 18th century? At first, it seems like a straight-forward process of closely studying the histories, treatises, reports and other materials that the historians left behind. The challenge, however, is that of scale. Studying one history closely or two comparatively offers different perspectives from studying 20 works produced by 11 historians computationally. Thus, in this poster, I present a section from my dissertation research where I computationally study a large corpus of Ottoman court histories from the long 18th century. I divided the poster along four thematic sections: historical introduction, natural language processing (NLP), findings, and methodological implications.

The archive of Ottoman histories at the center of this poster were written in the 18th and 19th centuries by court historians, who were also members of the Ottoman administration. Thus, they had their own positionalities with regards to the events they were recounting. Their narratives offer insights about the relationship between imperial rule and legitimacy through history-writing, particularly at a time when the Russian and Habsburg expansions were putting pressure on the Ottoman state from its borderlands. For example, the 1783 Russian Annexation of Crimea and the 1789 Napoleonic Invasion of Egypt feature prominently in these histories. By reading these sources computationally, I study general trends, discursive practices and the development of history-writing among multiple generations of the Ottoman elite. The questions prompting these analyses are informed by close readings of these sources and in turn, their results complement traditional historical methods. This section contains an introduction to the period with maps and a network visualization of Ottoman historians and their intellectual world.

NLP is an interdisciplinary field that deals with the computers’ capacity to process human language. From information retrieval systems (Google search engine), voice recognition technologies (Alexa) and text generating artificial intelligence (ChatGPT), many of our everyday interactions with computers rely on NLP. For historians NLP tools offer novel ways to develop arguments about archival sources. In this project, I use named-entity recognition and topic modelling to plot the central events, places, and actors in the court histories; experiment with cosine similarity to study how different historians copied sections of each others’ works, and explore the potential of building a semantic vector embedding system to offer glimpses into the conceptual scapes of Ottoman historians and the court. The findings in this poster will reflect the results of these NLP applications. This part contains details about the corpus, multiple plots and data visualizations.

Finally, the last section of this poster will reflect on the methodological and disciplinary implications of computational text analysis. With the release of ChatGPT, we find ourselves at a tipping point in our relationship with text. As historians, we cannot be passive recipients or blind critiques of these technologies. Thus, this poster places itself as an exploration into what computational text analysis can do for history.