Call for papers: Beyond Borders. Futures of Digitised Historical Media Collections for Research and Teaching

Historical media do not merely reflect the norms, ideas, and knowledge horizons of past societies — they actively shaped them. Historians of media have long aspired to study more than single institutions or single media types. Recent scholarship emphasises the dense relations between media — between newspapers and radio, for instance, marked by rivalry as much as by mutual learning, appropriation, and collaboration. More broadly, recent research focuses on the media ecosystems in which individual actors, institutions, infrastructures, and politics together shaped how media developed and functioned. The opportunities and challenges inherent in research using digitised collections are a meeting ground for both the historical and the digital humanities research communities.

For decades, the practical and methodological boundaries of research — between languages, source types, collections, and disciplines — kept such ambitions largely out of reach for empirical, source-driven work at scale. These boundaries are now in motion. Document processing that once demanded substantial expertise and resources has become widely accessible: text recognition, audio transcription, annotation, and classification no longer pose the barriers they once did. Even experimental techniques such as alignment across languages or source types, such as text, image or audio, are rapidly gaining traction. The obstacles have not disappeared — quality, coverage, and access remain uneven — but new techniques to work around them have emerged and with them the questions we can ask.

Beyond Borders invites researchers, cultural heritage professionals, and collection holders to reflect critically on the status quo and to think ahead: How should we conduct our research on and teaching with historical media collections in a time in which it has become easier than ever to connect them?

We invite submissions of original research, experimental methods, reusable resources (e.g. datasets, models, teaching materials) but also critical reflections and desiderata. We explicitly welcome contributions from scholars who do not themselves work computationally but whose work can constructively critique, inspire and inform the further advancement of data-driven approaches to the study of historical media.

Conference themes

1. Studying historical media ecosystems

The field of media history is widening its perspective beyond single media institutions and single media types, towards intermedial relations and entire media ecosystems. Computational research on historical collections is only beginning to catch up with these historiographical ambitions. We invite contributions which address these challenges in research and teaching:

• Case studies which analyse historical media ecosystems using traditional and computational methods
• Reflections on the critical analysis of different source types and their idiosyncratic logic
• Analyses of the differences and parallels in the evolution of historical media
• Methods to bridge the gap between non-digital, digitised and digital sources and research practices
• Best practices to teach (digital) research methods with digitised media collections

2. Embracing source heterogeneity

Historical research strives to weave together a narrative informed by the critical study of different source types. The goal is to produce comprehensive and carefully compiled analyses, which are based on relevant and available sources, take into account the positionality of the researcher and reflect the multiperspectivity and complexity of the historical record.

We observe different strategies to achieve this goal. Traditional historical research practices analyse highly diverse sources on a small scale at the expense of scale and breadth. This offers rich insights, for example, into the histories of single institutions, the impact of individual actors or evolving media formats.

Data-driven approaches typically centre on the analysis of a single source type at large scale at the expense of diversity and depth. Research focuses on identifying patterns within large, homogeneous datasets. This has widened the gaze, revealing patterns unobservable through close reading alone, but it has also significantly narrowed it, excluding relevant sources for lack of digitisation or incompatible data models. There remains a considerable gap between the traditional way of conducting research and what data-driven methods have so far made feasible.

One way to bridge this gap is the principle of scalable reading (shifting between close and distant reading) that has become a near-commonplace for homogeneous collections (books, newspapers, etc) but not yet for collections consisting of different modalities (e.g. text, audio, image). Recently, advances in multilingual and multimodal modelling demonstrate that the border between close and distant reading can also be crossed computationally. Models for images, audio, and video have shown that meaningful connections can be drawn across modalities. Embedding-based search – search based on vector representations of documents – offers a genuine alternative to reductionist keyword queries as it promises search for meaning rather than surface form.

We invite contributions which address these challenges, demonstrate best practices or challenge the status quo inspired by the following topics:

• Balancing the breadth in the study of historical media ecosystems demands with the depth that case studies require
• Methodological frameworks which enable comparative perspectives across national, linguistic, temporal or institutional borders
• Opportunities and challenges offered by generative AI, and more specifically, cross-lingual and multimodal embeddings for the presentation of empirical findings in historical narratives

3. Exploring human-machine interpretation

Distinct from, though related to, multimodal embeddings is the integration of large language models (LLMs) into historical media research. Here lies obvious potential to scale up practices traditionally associated with qualitative analysis: the careful, interpretation-heavy annotation and classification of sources. This shift — from probabilistic word counting to machine interpretation — sees computational tools taking on intellectually demanding tasks that were until recently the exclusive domain of the human researcher. LLMs also show a degree of robustness to poor OCR quality and offer workarounds for the persistent absence of optical layout recognition, two long-standing constraints on data-driven work with historical collections. The opportunities for efficiency are evident, and many experiments are ongoing, but many challenges remain:

• Integration of machines in research workflows in light of epistemological, methodological, ethical, technical, and legal challenges
• Impact of machines on human research designs
• Consistency, transparency, reproducibility and reliability of AI-assisted output
• Evaluation and human-gold standards to evaluate machine interpretation
• Reuse of models trained on historical media collections and insights gained from their creation

4. Connecting heritage collections

Historical media collections remain dispersed across institutions with diverse and evolving archival principles, metadata standards, and legal frameworks. Crossing these borders is as much an organisational and legal challenge as a technical one — and one that researchers and collection holders can only address together. The Collections as Data movement has achieved a paradigm shift in recent years, but important challenges and open questions remain:

• Connecting sources at scale across languages and modalities such as text, image, sound, and objects
• Evolving user needs and expectations
• Management of imperfection and critical assessment of the value and limitations of heterogeneous, unevenly documented collections, which are subject to divergent and constantly evolving archival policies
• Raising awareness of the significance of these challenges and sharing best practices to overcome them in teaching of digital literacy and digital source criticism

Submission formats

We welcome submissions in the following formats:

• Long papers: abstract of up to 1,500 words presenting original, completed, or substantially advanced research
• Short papers: abstract of up to 750 words presenting work in progress
• Posters: abstract of up to 500 words, including presentations of tools, infrastructures, datasets, and models

All submissions must include at least five academic references and must be written in English. Submissions will undergo double-blind peer review by at least two members of the programme committee.

Submit through EasyChair.

Timeline

• Submission deadline: 30 September 2026
• Author notification: 31 October 2026
• Camera-ready abstract deadline: 15 December 2026
• Conference: 28-29 January 2027

Presentation and participation

Accepted long papers will be allocated 20-minute slots and short papers 15-minute slots, including discussion time. All sessions will be livestreamed, and remote presenters may speak live by video link. Recordings will be available to registered participants after the conference.

Selected papers will be invited for publication in a peer-reviewed venue. Further information about the publication venue will be provided by 31 October 2026.

The conference will take place on the campuses of EPFL and the University of Lausanne. Registration details and fees, including reduced rates for early-career researchers and participants with limited travel budgets, will be announced by 31 October 2026. Online participation will be available for all sessions.

About the conference

This is the third and final conference organised by Impresso - Media Monitoring of the Past II: Beyond Borders: Connecting Historical Newspapers and Radio. It follows two project conferences devoted to transmedia relations in 2025 and the integrated history of radio and newspapers in 2026.

Since 2017, Impresso has developed tools and infrastructure for the computational analysis of historical newspaper and radio collections across languages and institutions. The project has worked on many of the questions addressed by this conference, including the connection of heterogeneous sources, the application of natural language processing to large multilingual corpora, and the design of interfaces grounded in historians’ research practices.

The conference will bring together researchers, teachers, developers, cultural heritage professionals, and collection holders working on these questions from different perspectives. It will provide a forum to assess current methods and infrastructures, identify unresolved needs, and discuss research and teaching practices suited to rapidly changing computational possibilities.

Organisers

On behalf of the Impresso Project: Marten Düring, Cao Vy, Caio Mello, Ferdaous Affan, Maud Ehrmann, Raphaelle Ruppen Coutaz, Kaspar Beelen

For enquiries, please contact info@impresso-project.ch.