Since 2025, the Faculty of Computer and Information Science (FRI) at the University of Ljubljana has been home to the European Centre of Excellence in AI for Digital Humanities (AI4DH). Funded by the European Union for five years, the project brings leading international researchers to a widening higher education institution, supporting structural change, strengthening research capacity and fostering brain gain. Dr Antoine Doucet, professor of computer science at La Rochelle University in France joined FRI UL as an ERA Chair holder. We spoke with Antoine about his research, his vision for AI4DH and his views on the recent technological advancements.

What is your research focused on?

I work with embeddings, transformers and large language models (LLMs), with a particular interest in historical documents. I have worked extensively with digitised documents. These are documents that were not originally created on a computer but were scanned at some point. When you scan a document, you lose much of its original structure, so you first need to reconstruct that structure and extract the text. This applies to both printed and handwritten documents.

The goal is to extract information from these documents and analyse large collections of them in ways that can help people make decisions. For example, banks may have hundreds of pages of PDF documents, but what they really need to know is something specific, such as a person's income or disposable income.

Historical newspapers are another good example. There are huge collections of newspapers, but analysing them presents several challenges. First, you need to identify and separate individual articles, which is not straightforward when you are working with scanned pages. Once you have reconstructed the documents, you can analyse their content at scale. For example, to understand who was discussing a particular topic, where the discussion was taking place and why.

You won the Europa Nostra award for the NewsEye project. Could you tell us more about the project and its significance?

Within the NewsEye project, we developed a digital investigator for historical newspapers. We worked with 11 partner groups, including three national libraries and four research groups. One of our main goals was to improve OCR, the process of converting scanned documents into machine-readable text and to identify and extract individual articles. We also extracted semantic information, such as the names of people, places and countries, so that this information could be identified and marked within the documents.

We then used this data for large-scale analysis and developed a personal research assistant that could generate summaries of selected newspaper collections. The tool also helped researchers build their own collections and analyse them.

At the same time, the project led to new developments in computer science, particularly in the analysis of noisy data. Most importantly, we were able to demonstrate that these tools could support humanities scholars working across different languages and disciplines, and addressing very different research questions.

Has your work always been this interdisciplinary?

Yes. As far back as I can remember, I have been working closely with researchers in the humanities, helping them address their research topics and questions using computational methods.

AI4DH and reflections on recent technological developments

Could you tell more about the European Centre of Excellence in AI for Digital Humanities?

Our ambition is to become a go-to place in Europe for humanities and social science researchers who want to explore how AI can support their work. Today, researchers have access to unprecedented amounts of data. Artificial intelligence can help us process these large datasets, uncover patterns that might otherwise remain hidden and gain new insights much more quickly. We see AI4DH as an important part of supporting this transformation. We offer a wide range of support, including access to computational infrastructure such as your FRIDA supercomputer, education and training, support with project proposals and research methodologies, and opportunities for research visits.

We are also developing a curriculum for an “Artificial Intelligence for Digital Humanities” Master's module, as well as courses at PhD level and micro-credential courses.

Most importantly, we want researchers to come to us with their specific needs and research questions. We can help them understand what is possible with AI, what is not, which approaches might fit their research, and where there are still research gaps that computer scientists at the Centre could help address. In this way, we hope to build not just a centre for AI research, but a place where computer scientists and humanities researchers can work together to develop meaningful new research.

There has been growing public scepticism and backlash against AI in the recent year. What are your thoughts on this?

With every new technology, we need to be careful about how we use it. Cars, for example, are extremely useful for getting from A to B, but they also cause many deaths through accidents and contribute to climate change through carbon emissions. Yet we still use them because we recognise their benefits and try to manage the risks. I think we should approach AI in a similar way.

At the AI4DH, we are not advocates of using AI simply because it is available. We advocate for using AI where it is genuinely useful and where its use is reasonable. That means considering questions of ethics, explainability and resource consumption, and, where possible, using open models. We want to avoid black-box systems that produce results without us understanding how or why they reached them, particularly when those results may not be correct.

Our goal is to help humanities researchers not to have AI do the humanities research. The name of the Centre is AI for Digital Humanities, not AI doing Digital Humanities. AI should be a tool that assists researchers. Part of our role is therefore to help humanities scholars understand which AI tools and approaches are appropriate for their research questions, and when AI may not be the right solution. We certainly do not want to simply promote the AI agenda of large technology companies.

What are your future plans for the European Centre of Excellence in AI for Digital Humanities?

So far, it has been only 16 months, but we have already initiated many different activities. Our main challenge now is to keep them going and make sure that they are developed and delivered well. We are already seeing a positive impact, and our priority is to make sure that this work can be sustained in the long term.

We are working on many different areas, from hosting visiting researchers and creating datasets and tools to organising training and providing support to researchers. We want the Centre to be an open and collaborative space, so researchers, students and anyone interested in our work are welcome to get involved.