German-language newspapers flourished during the era of transatlantic mass migration in the long nineteenth century, when approximately 6 million German-speaking people migrated to North America. Already in 1732, Benjamin Franklin had published the first German-language newspaper in the U.S.: Die Philadelphische Zeitung. However, the news organ folded after only a few issues. Hundred years later, transatlantic mass migration, technical innovations in print, increasing literacy rates and the growing interest of the public in information of a secular, sensational and political nature resulted in the widespread sharing of information. By 1890, German-speaking migrants published up to 1,000 newspapers across states turning the German immigrant newspaper into the most widely read and influential non-English newspaper. Large and small papers alike found a home in the United States, including the Westliche Post in St. Louis, Missouri, which promoted Joseph Pulitzer, a rising figure on the journalistic scene.
Editors like Pulitzer borrowed, selected, copied, and modified news content from across the country and the Atlantic and reprinted it in their own venues. In this way, due to insufficient copyright legislations and enforcements, technical innovations in steamship, telegraph systems, and print, texts traveled transregionally. But editors reprinted more than articles about events, statistics, or announcements. Readers of German-American newspapers were able to dive into pages where news from the homeland were printed directly next to nationalistic poems or advertisements of German beer products. Today, two centuries later, newspaper digitization projects such as Chronicling America by the Library of Congress bring different newspapers together and with their web-interfaces provide access to millions of newspaper pages ranging from urban newspapers in Chicago, Illinois, to rural areas such as Hermann, Missouri. However, such interfaces offer no innovative or advanced ways other than keyword search in addition to metadata filtering to detect text reuse. Which keywords to enter if one does not even know what they are looking for to find previously “unknown” texts?
In my PhD-project “Text Mining America’s German-Language Newspapers, 1830-1914: Processing Ger(wo)manness,” I use digitized German-language newspapers from the Chronicling America database and deploy text reuse detection algorithms developed in The Viral Texts Project and Oceanic Exchanges (1). With this method, I have uncovered approximately 500,000 snippets in total of recirculated texts in various newspapers from different states (2). Based on the term “Ger (wo)manness,” I question whether the history of the German-American press can no longer be written as a dominant male space created by successful editors, because there are numerous reprinted texts for, by, and about women: women appear as protagonists in hard news about suicides of young domestic servants, as authors of short stories about the sorrows of German migrant women in the U.S. and as promotional figures in advertisements of medical products to cure female weakness.
The abundance of reprinted texts, the diversity of genres, and their similarities as regards a sensationalized and commercialized narrative style led me to consider additional different options on how to proceed methodologically. Is there a computational way to first classify texts systematically into genres and subsequently use a mixed-methods approach of close and distant reading to examine the intersection of reprinting, genre, and gender? I added “wo” to Germanness to emphasize the high frequency of women’s representations in the dataset, but I put it into brackets to illustrate that further methodological steps are necessary to systematically recover their stories using the digital archive and identify different genres without having to spend years of reading all 500,000 texts. In this presentation, I will show how I use a human-in-the-loop (HITL) approach that leverages both human and machine intelligence to create machine learning models that automatically classify reprinted texts into different newspaper genres.
(1) Cordell, Ryan and David Smith. (2017). Viral Texts: Mapping Networks of Reprinting in 19th-Century Newspapers and Magazines. http://viraltexts.org; Oceanic Exchanges Project Team. (2017). Oceanic Exchanges: Tracing Global Information Networks In Historical Newspaper Repositories, 1840-1914. DOI 10.17605/OSF.IO/WA94S. Available at: osf.io/wa94s.
(2) Keck, Jana. (2021). Text Mining America’s German-Language Newspapers, 1830-1914: Processing Ger(wo)manness. Zenodo. https://doi.org/10.5281/zenodo.5518019
digitalhistory: Jana Keck: Encoding Text Reuse, Genre, and Gender in Nineteenth-Century German-American Newspapers, in: Digital History Berlin (Blog), erschienen am: 3. Januar 2022, Internet: https://dhistory.hypotheses.org/1040.