Digital resources in the Social Sciences and Humanities OpenEdition Our platforms OpenEdition Books OpenEdition Journals Hypotheses Calenda Libraries OpenEdition Freemium Follow us

Preprint published: NER4all or Context is All You Need


The Chair for Digital History is happy to announce that we published our first preprint on arXiv:

Torsten Hiltmann, Martin Dröge, Nicole Dresselhaus, Till Grallert, Melanie Althage, Paul Bayer, Sophie Eckenstaler, Koray Mendi, Jascha Marijn Schmitz, Philipp Schneider, Wiebke Sczeponik, Anica Skibba, NER4all or Context is All You Need: Using LLMs for low-effort, high-performance NER on historical texts. A humanities informed approach, arXiv (4.02.2025), DOI: https://doi.org/10.48550/arXiv.2502.04351

Abstract:
Named entity recognition (NER) is a core task for historical research in automatically establishing all references to people, places, events and the like. Yet, due to the high linguistic and genre diversity of sources, only limited canonisation of spellings, the level of required historical domain knowledge, and the scarcity of annotated training data, established approaches to natural language processing (NLP) have been both extremely expensive and yielded only unsatisfactory results in terms of recall and precision. Our paper introduces a new approach. We demonstrate how readily-available, state-of-the-art LLMs significantly outperform two leading NLP frameworks, spaCy and flair, for NER in historical documents by seven to twentytwo percent higher F1-Scores. Our ablation study shows how providing historical context to the task and a bit of persona modelling that turns focus away from a purely linguistic approach are core to a successful prompting strategy. We also demonstrate that, contrary to our expectations, providing increasing numbers of examples in few-shot approaches does not improve recall or precision below a threshold of 16-shot. In consequence, our approach democratises access to NER for all historians by removing the barrier of scripting languages and computational skills required for established NLP tools and instead leveraging natural language prompts and consumer-grade tools and frontends.

Shawn Graham: Do It Yourself Digital Golems. Experiments with various AI, neural networks, and other technologies for archaeology

The golem is a creature animated by the words in its head. Emerging from Jewish traditions, the words that did the animation had connections with divinity, but the golems themselves despite this spark of the divine toiled at mundane chores. But sometimes, the golems would get out of control, and become figures of terror. My none-too-subtle analogy is this: animated by the ‘words in their heads’, the anthropomorphized neural networks of contemporary ‘ai’ (and ‘agi’: artificial general intelligence) are similar figures, with similar connections to the divine (we are told: ‘agi’ will transcend us; ‘agi’ will bring about the singularity; ‘agi’ will make us all posthuman). In this talk, I will explore some of my experiments trying to teach these technologies ethically and critically. I will explore some of the ways I have been trying to raise my own digital golems for my mundane archaeological needs. This is a talk filled with failure and experimentation. But maybe, also, something useful.

„Shawn Graham: Do It Yourself Digital Golems. Experiments with various AI, neural networks, and other technologies for archaeology“ weiterlesen

Folgert Karsdorp & Mike Kestemont: The application of unseen species models in history: correcting registration bias in nineteenth-century police reports from Brussels

Researchers of the past – whether historians, literary scholars or archaeologists – depend on the sources that have stood the test of time. That sample of history is usually far from complete, however. There are numerous reasons for this, such as natural causes (e.g., fires or floods), decisions at the level of archival policy (what do we preserve and what do we not?), and biases in the formation of the archives themselves. Data representing lower classes were long considered less relevant, for example, and thus socioeconomic factors likewise play a role in the survival of sources. In a series of recent experiments, we have explored how statistical methods from ecology can help us identify such gaps and biases in our knowledge. Those methods all find their basis in “Unseen Species Models,” which were originally developed to estimate the number of unique species in an environment. Just as ecologists try to estimate biodiversity from an incomplete sample, we apply the models to incomplete historical archives to measure the actual cultural diversity.

„Folgert Karsdorp & Mike Kestemont: The application of unseen species models in history: correcting registration bias in nineteenth-century police reports from Brussels“ weiterlesen

Amanda Regan: Mapping the Gay Guides: Using Digital History to Explore LGBTQ Travel Guides, 1965 – 2005

Mapping the Gay Guides (MGG) relies on the Damron Guides, an early but longstanding travel guide aimed at gay men since the early 1960s. An LGBTQ equivalent to the African American “green books,” the Damron Guides contained lists of bars, bathhouses, cinemas, businesses, hotels, and cruising sites in every U.S. state, where gay men could find friends, companions, and sex. The online mapping project explores different dimensions of American gay life through time, from bars and nightlife, bookstores, cinemas, and churches. Users to the site are able to navigate and investigate the dynamic and sometimes disappearing nature of LGBT spaces over time.

„Amanda Regan: Mapping the Gay Guides: Using Digital History to Explore LGBTQ Travel Guides, 1965 – 2005“ weiterlesen

Moritz Feichtinger: Just a Little Bit? Historicizing Born-Digital Objects

The information age is intrinsically digital. Knowledge, culture, and communication are increasingly created, transmitted, processed, and stored in digital formats. Consequently, historical inquiry, particularly when engaging with the latter half of the 20th century and beyond, must grapple with digital artifacts. These artifacts, often termed “born-digital” objects, are created by and within digital environments. Their scale, diversity, and inherent technical and cultural characteristics present distinct challenges for historical research, requiring historians to adapt established methodologies, particularly source criticism, to address these nuances.

„Moritz Feichtinger: Just a Little Bit? Historicizing Born-Digital Objects“ weiterlesen

Studentische Veranstaltung: DH-FuckUp-Outing 2024/1

TL;DR: Donnerstag, 19.12., ab 18 Uhr c. t.;
Input, Austausch und Snacks; Anmeldung formlos via Mail.

»Ohne die richtige Einstellung ist erfolgreiches Scheitern unmöglich. Man muss Scheitern wollen – sonst wird es nur eine peinliche Niederlage.«
Alexander Florin, 2013: Erfolgreich Scheitern (via DWDS-Korpussuche)

Erzählen vom und Sprechen über das Scheitern – das ist die Idee hinter dem Konzept ‘FuckUp-Night’.1 Über die Kommerzialisierung dieser Idee und den Grund für den legally distinct Namen dieser Veranstaltung lässt sich streiten; das Konzept hat sich aber bereits seit mehr als zehn Jahren bewährt.
Dem gemeinsamen Sprechen über Scheitern wohnt eine radikale Kraft inne: Wir alle scheitern im Großen wie im Kleinen – meist eine sehr einsame Erfahrung. Aber genau diese Einsamkeit gilt es gemeinsam zu durchbrechen!

„Studentische Veranstaltung: DH-FuckUp-Outing 2024/1“ weiterlesen

Wonhee Cho: Rethinking ‘Sinicization’ through Data: A Network Analysis Methodology for Yuan Dynasty Studies

This presentation revisits the concept of “Sinicization” during the Yuan dynasty by employing network analysis to evaluate the centrality and cooperative dynamics of officials in Yuan court. Using data derived from the Official History of the Yuan (Yuanshi), this research interrogates the validity of traditional narratives of cultural assimilation, particularly the extent to which the Mongol administration in China adopted Han Chinese governance practices. By focusing on the relationships among central officials, this study sheds light on the evolving power dynamics between Han and non-Han elites under Mongol rule.

„Wonhee Cho: Rethinking ‘Sinicization’ through Data: A Network Analysis Methodology for Yuan Dynasty Studies“ weiterlesen

Ashley Sanders: Redressing Colonial Knowledge Systems through Restorative Data Justice

While “data” is often understood today in computational terms, as information coded and organized for interpretation with digital tools and algorithms, the term has a long history dating back to at least the seventeenth century. In its earliest uses, data was defined as a “given” and a basis for decision-making—and thus power. A study of colonial knowledge systems offers numerous examples of the link between power and data regimes. Revealing the integral role of data in the building and maintenance of empires, this talk takes up several methodological questions: How do we handle colonial data, both data generated by colonial administrations and data we recreate from fragmentary sources to address the absences in the historical record born from conquest? And how do we ensure that our own research does not replicate imperial extractive colonial data practices?

„Ashley Sanders: Redressing Colonial Knowledge Systems through Restorative Data Justice“ weiterlesen

Clemens Neudecker: Context matters. Chancen und Herausforderung bei der Arbeit mit Künstlicher Intelligenz und Kulturellem Erbe

A banana on a table and an image of a banana on a laptop on the same table. Each of the two bananas has a white frame around it with the word ‚Banana‘ sticked on top of it

Für Bibliotheken bieten die auf dem Gebiet der Künstlichen Intelligenz (KI) erzielten Fortschritte vielfältige Chancen. Die Staatsbibliothek zu Berlin – Preußischer Kulturbesitz (SBB) entwickelt im Rahmen von Projekten wie Mensch.Maschine.Kultur KI-Technologien für eine Vielzahl von Anwendungsbereichen: von Text- und Layouterkennung über Bildanalyse bis zu maschinell unterstützter Sacherschließung und nicht zuletzt Datenbereitstellung und digitaler Kuratierung. Andererseits sind bei der Anwendung von KI-Technologien für das Kulturelle Erbe immer auch die historischen und kulturellen Kontexte zu beachten.

„Clemens Neudecker: Context matters. Chancen und Herausforderung bei der Arbeit mit Künstlicher Intelligenz und Kulturellem Erbe“ weiterlesen

Shih-Pei Chen: Treating a genre as a knowledge system: Analysing Chinese local gazetteers and daily-use encyclopedias with quantitative approaches

Many digital humanities research projects rely on large amounts of digitized primary sources to elevate research from manual, close reading to large-scaled, computer-assisted quantitative analyses –– sometimes called distant reading or birds-eye views. This presentation reports on the case of Chinese local gazetteers (difangzhi), in particular the Local Gazetteers Research Tools (LoGaRT), designed and developed at the Max Planck Institute for the History of Science (MPIWG) to bring a collective, quantitative lens to research the genre.

„Shih-Pei Chen: Treating a genre as a knowledge system: Analysing Chinese local gazetteers and daily-use encyclopedias with quantitative approaches“ weiterlesen