The paper illustrates how the combination of frequency analysis and diachronic word embeddings can help to explore and understand the development of long-running discourses in large sets of historical data.

The paper shows how combining frequency measures with diachronic word embeddings helps trace long-running political discourses in large historical corpora, revealing how meanings and associations shift over decades.

To assess the semantic changes in how German politicians spoke about the Wehrmacht, we use overlapping diachronic word embeddings. An additional analysis of term frequency measures the level of interest in the Wehrmacht over time.

The Bundestag - (West) Germany’s federal legislature - has published its proceedings since 1949. Digitized editions of these proceedings contain over 900,000 speeches, questions, and interjections, forming a valuable but underutilized corpus of sources on post-war German politics.

Debates on a Burden: Exploring the Connotations of the Wehrmacht in Post-War German Politics This paper investigates the evolving connotations of the Wehrmacht in post-war German parliamentary politics. By analysing term frequencies and comparing diachronic word embeddings on a corpus of German parliamentary debates, we measure prevalence and context of references to the Wehrmacht over time. The results show that the Wehrmacht went from being a prominent issue of defence and welfare policies to near irrelevance by the 1960s. They also indicate that the resurgent post-reunification controversies on the Wehrmacht’s war crimes did not mark a radical shift in political culture, but followed a decades-long trend of increasingly strong associations between the Wehrmacht, war crimes, and Nazism in political speech, that developed alongside an emerging culture of remembrance.

How have German politicians dealt with the burdensome legacy of militarism after the Second World War? This paper explores the connotations of the Wehrmacht in political speech through word embeddings trained on parliamentary proceedings.

Many thanks for the team at @jdighist.bsky.social for everything to bring this article to fruition! Check it out to learn more the utility and limits of LLMs as digital tools for historical research - like using LLMs for OCR, which is a hot topic of late.

How did early modern people imagine the future—and who shaped it? Analyzing 4,900 English letters (1410–1695), Sara Budts shows how phrases like “God willing” slowly faded, revealing a shift from Providence toward human agency and a more uncertain, self-shaped future.

What can 7,500 divine references in early modern letters tell us about how people imagined the future? Using the CEEC, Sara Budts maps when and how writers invoked God, revealing a patchy decline: routine phrases faded, but appeals persisted in illness, crisis, and war.

From “God willing” to “I trust in God,” early modern letters show how people used divine appeals to navigate uncertainty. Through corpus analysis, Sara Budts traces England’s uneven secularisation: routine invocations faded, but personal faith deepened, especially in times of crisis.

The secularisation of future expectations in practice: An empirical study of divine appeals in Early Modern English letters In the wake of Reinhart Koselleck’s seminal work on temporality (1979), historians studying past futures in Western Europe have argued that our current understanding of the future dates back to the period between 1500 and 1800. The medieval, Christian conception of time was largely cyclical in nature; the future was, above all, in the hands of God. By 1800 however, the future had become open, uncertain and constructible; people were left with the feeling that time had not only been accelerating, it had also become secularised. As recent studies have emphasised the gradual nature of this shift, this paper zooms in on the pluritemporal mindscape of early modern societies by charting secular and religious types of future thought in a large body of English letters written between 1450 and 1700. Did fifteenth century people appeal to God more often than seventeenth century people did? In which domains of their lives was religious future thinking the strongest? Did the secularisation of time proceed at the same pace across all communities, despite their differences in religious practice? We address these questions by querying nearly 5000 early modern letters for divine appeals and systematically annotating them for variables like human and divine agency, temporal orientation and domain of life. Our results indicate that while the more formulaic divine appeals found in the opening and closing sections of letters were growing less popular over time, the ones in the letter bodies fluctuated in particular with the religious denominations of the letter writers. The observed rise in mentions of divine entities in the first half of the seventeenth century is mainly caused by a small group of puritan letter writers whose involvement in the civil wars throughout the 1640 made their lives particularly perilous. The other letter writers in the corpus, by contrast, displayed progressively lower rates of divine appeals as time went by, a finding that is in line with previous research that saw the early modern period as one characterised by the increasing secularisation of future thought as well as a shift from religious practice to religious faith.

Early modern letters are full of phrases like “God willing” or “By God’s grace.” Sara Budts’ analysis of 5,000 letters shows these weren’t clichés but ways to navigate faith, agency, and uncertainty. From 1450–1700, people balanced divine will with human action in shifting ways.

From tools to sources: By examining LLMs through source criticism, historians can evaluate generative AI's uneven potential while contributing to broader debates about these technologies' societal impact.

Mapping AI's historical knowledge: Benchmarking LLM performance on historical facts, oral history transcription accuracy, and OCR correction tasks reveals uneven capabilities. Results show whose history has been digitized and made computationally legible. #oralhistory #OCR #benchmarks

Mapping the Latent Past: Assessing Large Language Models as Digital Tools through Source Criticism This article examines how digital historians can use large language models (LLMs) as research tools while critically assessing their limitations through source criticism of their underlying training data. Case studies of LLM performance on historical knowledge benchmarks, oral history transcriptions, and OCR corrections reveal how these technologies encode patterns of whose history has been digitised and made computationally legible. These variations in performance across linguistic and temporal domains reveal the uneven terrain of knowledge encoded within generative AI systems. By mapping this "jagged frontier" of AI capabilities, historians can evaluate LLMs not just as tools but as historical sources shaped by the scale and diversity of their training. The article concludes by examining how historians can develop new forms of source criticism to navigate generative AI's uneven potential while contributing to broader debates about these technologies' societal impact.

Examining LLMs as historical sources: This study applies source criticism to AI training data, revealing how patterns of digitization shape what these tools encode. Case studies map this 'jagged frontier' of capabilities across historical tasks, languages & time periods. #DigitalHistory #LLMs

History-focused: What happened to Swedish public service radio when commercial broadcasting arrived? Analyzing 1,600 hours of P1 & P3, this article uncovers shifts in sonic patterns & content variation, offering new insights into the sound history of broadcasting

Navigating the Radio Archive: Segmentation and analysis of Swedish broadcasting data This work aims to explore the relationship between sound archives and historiography, focusing on the Swedish case of mass media archiving from the 1980s. The study investigates the sonic content of public service radio changes during the introduction of commercial broadcasting, using computational methods. In the analysis, two of the most popular public service broadcasting channels, P1 and P3 are compared. By analysing 1,600 hours of radio data, the paper reveals a shift in the number of detected occurrences, with varying sonic sequences that reflect the overall structure of the broadcast. Although the sample sizes are small, the findings show a correlation between object detection and dimension reduction, suggesting and increasing attention to content variation. The paper contributes to the understanding of historical radio data, offering pre-processing and segmentation methods for working with cultural audio data. It also emphasizes the methodological implications of combining dimensionality reduction and object classification approaches, demonstrating the value of using pretrained and untrained algorithms together for a comprehensive understanding of the local and fine-grained aspects of audio data. The necessity of such an approach springs from the oceanic extent of content in the radio archive. Thus, the article suggests new ways to navigate the radio archive

Methodology-focused: How can historians work with thousands of hours of sound? This article introduces segmentation & dimensionality reduction methods for large-scale radio archives, showing how computational analysis reveals hidden structures in public service broadcasting

Journal of Digital History The Journal of Digital History (JDH) is an international, academic, peer-reviewed and open-access journal. JDH will set new standards in history publishing based on the principle of multi-layered articles.

The research shows how digital tools like Aventura.js can mediate archival inquiry. Through an interactive atlas and the Carguero case study, it offers a new approach to reading visual history. journalofdigitalhist...

Inspired by Warburg’s principles, the article transforms static images into interlinked, interactive panels. Its case study of the 19th‐century Carguero deepens our understanding of visual symbolism.

By merging digital techniques with human interpretation, the article demonstrates how interactive panels and data visualizations offer fresh insights into visual archives via the Carguero case.

The study repurposes Aventura.js—a bilingual, open‐source tool—to build digital atlases that trace symbolic connections in historical images, exemplified by Colombia’s Carguero.

Journal of Digital History The Journal of Digital History (JDH) is an international, academic, peer-reviewed and open-access journal. JDH will set new standards in history publishing based on the principle of multi-layered articles.

A forthcoming article by Rodríguez Gómez and Urueña presents a method for exploring visual collections using interactive panels, inspired by Warburg’s Mnemosyne Atlas, to analyze the 19th‐century Carguero case. journalofdigitalhist...

But these sources raise questions. Can we ethically use archived posts from forums & blogs without consent? How do we avoid overinterpreting what's preserved—and what's lost? The paper ends with a call for cautious, critical digital history.

From lean employment pressures to restrictive benefits, the 2000s marked a turning point. Web archives reveal raw, everyday reflections—from sick workers to HR managers—offering insight into how people understood these changes, not just endured them.

Journal of Digital History The Journal of Digital History (JDH) is an international, academic, peer-reviewed and open-access journal. JDH will set new standards in history publishing based on the principle of multi-layered articles.

How did British citizens experience sickness & disability policy under New Labour? This article uses blogs, message boards & the Internet Archive to explore how workers, managers & families navigated welfare reform—and how they remembered the welfare state. journalofdigitalhist...

GaLiLeO is unfinished & its corpus incomplete. But that’s the point. It shows how partial, purpose-built tools can still yield insight—especially when they trace relationships that archives and metadata might miss. A third space between archive and algorithm.

Unlike most text mining tools, GaLiLeO doesn’t aim for sweeping models. Instead, it helps researchers identify meaningful links across Galileo’s library—prioritizing interpretive, humanistic inquiry over scale. It’s built for discovery, not datafication.

Journal of Digital History The Journal of Digital History (JDH) is an international, academic, peer-reviewed and open-access journal. JDH will set new standards in history publishing based on the principle of multi-layered articles.

What if digital tools didn’t summarize a whole corpus—but helped historians find meaning in just a few documents? 📜 GaLiLeO (Galileo’s Library and Letters Online) is a prototype digital lab that builds interpretive paths through Galileo’s letters, books & notes. journalofdigitalhist...

Journal of Digital History The Journal of Digital History (JDH) is an international, academic, peer-reviewed and open-access journal. JDH will set new standards in history publishing based on the principle of multi-layered articles.

We worked to recover and make accessible 10,000+ defaced websites, captured between 1995–2010. These archived hacks aren't just noise - they reveal political messages, digital protest, and forgotten web aesthetics. Here's how they help us understand web history: journalofdigitalhist...

Journal of Digital History The Journal of Digital History (JDH) is an international, academic, peer-reviewed and open-access journal. JDH will set new standards in history publishing based on the principle of multi-layered articles.

What can hacked websites tell us about the history of the web?
From online activism to digital vandalism, political web defacements were part of shaping the early internet. We just published a new article in the JDH diving into this fascinating and underexplored archive. journalofdigitalhist...

Excited to introduce Arvest - the platform which will be at the core of the ERC 'From Stage to Data’ project. Designed to process and analyse digital traces of performing arts, it's transforming how we study theatrical heritage. Explore arvest.app #PerformingArts #DigitalArchives"

Discover Arvest, the cloud-based platform for IIIF image & video annotation. No server installation needed - just create an account and start annotating your digital resources immediately. Try it today at arvest.app #IIIF #DigitalHumanities #Annotation journalofdigitalhist...