CRASH: A Hackathon for Advanced Search in Cultural Heritage

Search is at the heart of research. Many decisions of a research tool come together in its search: what counts as similar, why a result is shown, who can use the tool, and what a query costs. For two days, CRASH, the Cultural Heritage Advanced Search Hackathon, puts cultural heritage data and the people who care about it in one room: on 30 November and 1 December 2026 in Graz.

About the Project  

At CRASH, teams of two to five people build search prototypes on real material and test them. The motto is “Vibes and Checks”: teams build freely, with their own data and problems, and each team picks one focus area in which it checks its tool at the end. The room chooses the winner of hearts, the jury the winner of numbers. The five focus areas are:

• Efficiency: what does one search cost in time, compute and energy?
• Trust: can I see why this result is here?
• Control: can the user steer what “similar” means?
• Access: who can use it, with a screen reader, by voice, in plain language or in a second language?
• Discovery: does the search find the unexpected, measured beyond accuracy?

Topics include retrieval over image, text and layout at the same time; embeddings and what they miss; re-ranking; retrieval-augmented generation that cites its sources; agents that keep a log of what they did; search across languages, scripts and historical spellings; detection of copies and near-duplicates; recommendation and exploration instead of a query box; knowledge graphs and linked open data; evaluation beyond accuracy; and the cost of one query. Two impulse talks, one by a domain expert and one by a technical expert, frame the first day.

We invite students, researchers, practitioners and experts from academia, memory institutions and industry. Coding helps, but not everybody needs it: participants bring a skill, a dataset or a problem they know well. Taking part is free of charge.

Data and infrastructure

Teams bring their own datasets. The organisers provide prepared datasets on Austrian cultural heritage. The DHinfra clusters in Graz provide current open-weights models and GPUs. The ecological footprint of the event, in CO2 and compute, is measured and published. Travel and accommodation are supported on request.

Organisation

CRASH is organised by the Department of Digital Humanities of the University of Graz, Graz University of Technology and Know Center.

Results

Working prototypes on real material, each with a test that shows what it does. After the event, the CRASH page shows the teams and what they built. Every team is asked to share its code under Apache-2.0 or MIT.