‘As digital methods prove useful, more scholars become open to what’s possible’

17 November 2025

Renowned researchers regularly come to the CSMC as Petra Kappert Fellows. Elisa Barney Smith is the first computer scientist in this programme, which she participated in for the second time this fall. In our interview, she talks about the growing connection between her field and manuscript research.

Elisa Barney Smith, you have been doing research on historical document analysis for over 30 years. Can you give us an idea of how the field has developed in this period?

When I started, the technology was quite limited. Storage and computer processing power weren’t what they are today. Early research primarily focused on segmenting and recognising typewritten or typeset texts, usually in black-and-white and mostly in English. As computers became more powerful, our field expanded. We moved from printed documents to all kinds of handwriting, then to colour images, and eventually into scripts beyond the Latin alphabet as well as different writing materials. Today, document analysis addresses complex historical challenges, including manuscripts written on palm leaves, palimpsests, documents with bleed-through from ink, and much more.

This fall’s fellowship at the CSMC was your second stay at the centre, following an initial visit in the summer of 2024. How did you first become involved?

My initial connection with the CSMC came through Hussein Mohamed, whom I first met at a Document Analysis Summer School in Greece. He was then a doctoral researcher, and we kept in touch, meeting regularly at conferences like the International Conference on Document Analysis and Recognition (ICDAR). Our collaborations picked up momentum after I moved from Boise State University in the US to Luleå Technical University in Sweden. Hussein, now a Principal Investigator at the CSMC, encouraged me to apply for the Petra Kappert Fellowship, and that’s how I first landed in Hamburg. It’s been invigorating to be in such an interdisciplinary environment, meeting people passionate about manuscripts and working on truly unique problems.

What did you work on specifically during your time here?

I got engaged in several different projects, for instance on the analysis of historical paper. Malgorzata Grzelec and Agnieszka Helman-Wazny investigate how different historical techniques of papermaking affect the alignment and structure of paper fibres. Samples were reproduced using various methods, and I am developing code to analyse 3D computed tomography scans of these paper samples. By identifying patterns in the fibre orientation, we hope to build a ‘fingerprint’ for each manufacturing technique, potentially allowing us to date and localise historical manuscripts based on their materials.

In addition, since my last visit I developed algorithms to extract strokes and features from handwritten scripts, supporting research into how physical conditions influenced the development of writing systems, which we are working together to modify to fit their research needs.

I also had the opportunity to talk to Mark Soileau, who is studying Ottoman Turkish manuscripts and wants to compare different editions of specific texts. My team in Luleå is showing great success in automatically transcribing Arabic-family scripts. Having transcriptions will help us analyse differences between editions to answer the research question: Are certain deviations due to mistakes by the scribe or are they deliberate? My experience from working with English and Latin documents, including aligning different editions of historical texts, proved quite relevant.

There’s a myth that scientists and engineers aren’t interested in history or the arts, but many of us do this work because we’re curious about culture and humanity’s past.

How do your collaborations with humanities scholars typically begin and develop?

Very often, they happen by serendipity. My first interdisciplinary collaboration started at Boise State University about 15 years ago, where a colleague from the English department was interested in applying optical character recognition to pages of printed books once owned and read by 19th-century author Herman Melville, and I realised that I can help him with that. We were able to get a small amount of seed funding from an organisation on campus, hired a master student, and built a system that he was able to use. Long-term partnerships often develop out of small projects like this one. It takes time to get to know each other, learn to speak a shared vocabulary, and build trust. Since then, I have regularly sought out opportunities to collaborate, whether through university initiatives in France or Germany, or through personal connections at conferences or workshops.

Do you sense a shift in how humanities researchers approach technology and digital methods?

Yes, especially at places like CSMC. Of course, it’s challenging to build interdisciplinary teams. Everyone already has full commitments and established projects. But as digital methods prove useful, I notice more scholars becoming open to what’s possible. Even if our first collaborative projects are informal or experimental, seeds are planted for deeper, grant-funded research in the future.

What about the future prospects of the work you have started with researchers here at the CSMC?

One exciting area is automatic recognition of handwritten Arabic-script documents, relevant to several ongoing projects at the CSMC, including work with audition certificates in Arabic. I’m now co-supervising a PhD student in Luleå whose research focuses on this very topic, and we hope that her techniques will advance the work that is done in this area at the centre.

In addition, work on Ajami manuscripts, written in African languages using Arabic script, has great potential. We were able to improve upon the work reported on a dataset from a colleague in Uppsala, moving text recognition accuracy from 65–85% error rates with existing tools to below 10% error with our methods. That opens up practical applications like automatic summarisation or translation of historical manuscripts and will help them better index their collection.

There are also promising opportunities for collaboration using data from the Deutsches Literaturarchiv, where CSMC researchers have already developed tools to annotate a large amount of images, which can be used as input to our text line recognition algorithms. Improving the analysis and categorisation of these sources will help add richer metadata, enabling new forms of research for literary scholars and historians.

What do you personally take away from these interdisciplinary exchanges?

Beyond the technical challenges, I genuinely enjoy the intellectual and cultural exchange. There’s a myth that scientists and engineers aren’t interested in history or the arts, but many of us do this work because we’re curious about culture and humanity’s past.

I’ve always loved travel and learning about the places I visit, Hamburg included. The city itself pleasantly surprised me; despite its size, it feels welcoming and manageable. If the winds of fate had blown me differently, I easily could have moved there.