What do an ancient coin, an oil lamp from the time of Jesus and a travel photo of a mosque have in common? Whilst researchers and anyone interested previously spent long hours of deliberation, overlooking parallels, the newly developed app “Tin Gustaf – Your Holy Land AI Pocket Museum” immediately reveals similarities with just a few clicks.
The app is based on automatically generated, English photo descriptions of artefacts from the Dalman Collection. These bear witness to the cultural landscape of Palestine around 1900: from historical architectural and portrait photography to pressed plants in a herbarium, agricultural machinery, and musical instruments. After clicking on an individual photo, further information such as photo motif, age and photographer are displayed as they are stored in the German Digital Library.
If you click on an artefact in the app, AI image descriptions are displayed that have been stored as data sets. The system uses two AI models to create these descriptions: the first model transforms every photo into text. The second model “translates” these texts into numbers (so-called numeric vectors) in order to separate the images into groups. The results are then combined in the app. “The Dalman App makes it possible for our collection to fit inside your pocket,” enthuses PD Dr. habil. Karin Berkemann, Curator of the Dalman Collection. “We now finally also have photo descriptions in English. This will profit users from the US, England, Israel, and Palestine, who use our collection for research.”
App reveals cross-references
The app enables new cross-references. Where researchers are often stuck in their own areas of specialisation, AI is able to recognise similarities without bias: “An ancient coin, an oil lamp from the time of Jesus and a travel photo of a mosque from the 1960s are displayed in the same group of images – because they all show the same ornaments that are described in the AI text,” explains Dr. Philipp Adämmer in interdisciplinary exchange with Berkemann.
As the app provides a fast and easy introduction to the collections, it is also well-suited for use in teaching. There are also plans to use the app in the Dalman Institute’s next exhibition in order to reach a broader target group, i.e. members of the general public who are interested in different cultures. The app can be used as inspiration and as a model for making other collections digitally accessible.
The app is based on open-source models. What is new is that the app combines AI applications for images and text: “This means that users receive more than just an individual photo description. They can also access other photos within semantically similar groups of images,” explains Karin Berkemann.
The Dalman Collection
The AI evaluates image data that has been created in the last 20 years during the digitisation of the Dalman Institute together with Greifswald’s Kustodie (University Collections). The initiator of the collection, theologian Gustaf Dalman (1855-1941) traveled to Palestine for the first time in 1899. In the subsequent decades, he researched how farmers and the Bedouin lived. His recorded observations include how they dressed, how they built their tents and houses, and how they looked after their guests.
The Dalman Collection has been catalogued at the University of Greifswald for more than 100 years. The artefacts are used for research, teaching, and educational work. They include historical photographs, plant samples, Bedouin clothing, agricultural machinery, ancient coins, and maps. The approximately 20,000 photos are particularly valuable and include diapositives, negatives, prints, paper copies and numerous aerial photographs from around 1900. They document the cultural landscape of Palestine prior to the First World War in a form that is unique to Europe.
Further information
https://dalman.nube.uni-greifswald.de/
Contact at the University of Greifswald
PD Dr. habil. Karin Berkemann
Curator of the Gustaf Dalman Collection
Faculty of Theology
Am Rubenowplatz 2-3, 17489 Greifswald
Tel.: +49 3834 420 2546
berkemannkuni-greifswaldde