Art DATIS: project update

Sybren Valkema was a prominent glass artist and teacher, born in 1916. Now, more than 100 years later, researchers working on the Art DATIS project aim to make his archives accessible to fellow researchers, artists and craftsmen. We sat down with Vera Provatorova (University of Amsterdam) and Carlotta Capurro (Utrecht University) for an update on this fascinating project.

Who was Sybren Valkema?

Carlotta Capurro is an art historian who likes working on the intersection of cultural history and the digital world. In previous research, she focused on the way digitalities impact the way we perceive culture and the role of (meta)data in sharing cultural heritage. She enthusiastically explains the significance of Valkema’s work: “Sybren Valkema was a glass artist working at the Royal Leerdam factories, a main producer of design glassware. When attending a seminar in the United States, he met a fellow glass artist that had developed the idea that working with glass should be available for individual artists, not just for people working at factories. Together with scientists, they developed a portable oven for melting glass and started reworking traditional knowledge into courses on art academies, to popularize working with glass as an art material. The importance of Valkema’s work is that he connected the European and the American traditions and innovated from that connection.”

The art of digitizing archives

Vera is a data scientist with a knack for interdisciplinary research. She has always felt it important to apply data science in a way that is meaningful to others. “The interesting thing about Art DATIS is that it combines data science and cultural heritage. If we can make Valkema’s archives accessible and searchable, this will be of great help to art historians and glass artists.”

However, making an archive digitally accessible is no small feat. It all starts with scanning every single page, Vera explains. “Picturae, one of the partners in this project, scanned the whole archive. While this has been instrumental to the project, scanned pages are stored as images, and turning them into text is another challenging step. Parts of the archive consists of typed text, which is relatively straightforward to work with. Most of the archive is hand-written however, and thus much more challenging. We use a model that has to be trained on handwriting of a particular person.”

Structuring the data

“Conversion of (hand-written) text from scans is just the start,” Vera continues. “The second challenge is that text will still be unstructured data after conversion. We decided that an approach called entity linking may be the best way to tackle this. In brief, we train an algorithm to recognize specific entities in a text (for example ‘Valkema’ or ‘Amsterdam’) and then we link that entity to a knowledge database. A particular challenge is to eliminate ambiguity: ‘Valkema’ could refer to the glass artist, but also to one of his relatives.”

“The art historians in this project are involved in the digitization process from an early moment,” adds Carlotta. “Interdisciplinarity is the real strength of this project. We aim to structure the data in such a way that is not only functional to Art DATIS, but in the most extensive way possible so that other researchers can benefit from it. For us it is very important to have the information available in such a way that we can reliably compare it to other sources. This can be challenging as there is not one international standard to structure semantic data – a broad challenge in our field that needs work.”

Asking Sybren

The Art DATIS team not only aims to answer their own research questions, but also works to make the archive available for other scholars. With RKD Netherlands Institute for Art History on board as a project partner, the aim is to make the archive available through their interface. Carlotta: “What I like is that we are working on the data without shaping the data. This makes it available for other researchers now and in the future.”

“Contemporary knowledge is often already somewhat structured, but for historical information this is not the case,” she continues. “The continuous challenge in projects like this is bridging unstructured knowledge and structured knowledge. Our contribution to solving this challenge is also what makes our work relevant to other projects in the cultural heritage field and even other research fields.”

Looking to the future, it would be great if the Art DATIS project could contribute to building a question answering algorithm, meaning that you can ask a question like ‘Where was Sybren Valkema born?’ and it will automatically retrieve the relevant information and formulate an answer. This idea was brought up by Anna Carlgren, a glass artist and former student of Valkema. The idea is outside the scope of the current project. “It would be challenging,” explains Vera, “because we are dealing with natural language processing on two sides - the question and the archive. But it would be amazing if we could build a virtual assistant like Alexa or Siri.”

While a virtual assistant called Sybren still requires a lot of work, the progress in the Art DATIS project is promising. For more information on the project, go to https://artdatis.nl/.