The way we interact with information is undergoing a significant transformation, and Large Language Models (LLMs) are at the forefront of this revolution. From internal knowledge bases to external, customer-facing documentation and support, LLMs are changing the game. However, the fast-moving landscape of AI is sometimes difficult for non-specialists to navigate. We therefore provide an overview of the latest developments in this field.

We are thrilled to invite you to the AI-on-Demand Showcase for Cultural Heritage. This event is aimed at professionals and researchers interested in the application of AI in cultural heritage, but should be of interest to anyone working with data.
The program will begin with an introduction to the AI-on-Demand platform, highlighting its role in advancing European AI research and innovation with a focus on quality and trustworthiness. Edisa Lozić will provide a brief overview of its relevance for the cultural heritage sector.
Following this, Benjamin Štular and Edisa Lozić will discuss the impact of Large Language Models and Retrieval Augmented Generation on data management in cultural heritage, supported by relevant case studies. Ronald Harasim will address the application of AI in the archival process, focusing on the enhancement of data processing and storage through technologies like OCR and image recognition. Additionally, Žiga Kokalj will introduce a tool designed for the Automatic Detection of Archaeological Features using machine learning, demonstrating its ease of use and minimal training requirements.
This event will offer a platform for learning about AI applications in cultural heritage using AI-on-Demand platform and for networking with experts in the field.

Agenda and detailed description

Benjamin Štular & Edisa Lozić
Data Conversations Made Easy: an overview for cultural heritage
The way we interact with information is undergoing a significant transformation, and Large Language Models (LLMs) are at the forefront of this revolution. From internal knowledge bases to external, customer-facing documentation and support, LLMs are changing the game. One technique that is gaining traction is Retrieval Augmented Generation (RAG), which is taking LLMs to the next level.
In this presentation, we will introduce you to the state of the art in how we can use AI in our interaction with data. Using three case studies, we will demonstrate how we can deal with structured, semi-structured, and unstructured data. All case studies are from the cultural heritage sector, but the concepts are generally applicable.
The focus will be on no-code methods that can be easily implemented even without prior knowledge of AI or application development. We also introduce BmyRAG, a tool that provides the user with a personalised roadmap to help them get started with the development of a RAG application.

Ronald Harasim
The Use of Artificial Intelligence for Processing and Analyzing Archival Data
The current development and gradual increase in archaeological data presents a challenge in their effective processing and archiving. We focus on the use of modern technologies, such as artificial intelligence, to streamline this process. Technologies like optical character recognition (OCR), natural language processing, image recognition, and other algorithms now enable the conversion of large volumes of diverse data into clear and easily searchable databases.

The presentation includes a brief general overview of the technologies used, with the main focus on demonstrating a part of the AIS ČR system, which will showcase the possibilities of using these technologies in practice. It will show how these tools can reduce the amount of manual work required to process data and increase the speed and efficiency of their further processing.

Žiga Kokalj
ADAF – a user-friendly ML tool for Automatic Detection of Archaeological Features
The need to use machine learning (ML) in archaeology is constantly increasing with the rapid development of image analysis techniques and the increasing availability of high-quality airborne laser scanning data (ALS, lidar). The tool for Automatic Detection of Archaeological Features (ADAF) has been developed to provide user-friendly software that uses ML models (in particular convolutional neural networks) to enable the automatic detection of archaeological features from ALS data. The software requires minimal interaction and no prior user knowledge of ML techniques, greatly improving its accessibility to the archaeological community. The underlying ML models have been trained on an extensive archive of ALS datasets in Ireland, labelled by experts with three types of archaeological features (enclosures, ringforts, barrows). The core components of the tool are the Relief Visualisation Toolbox (RVT) and the Artificial Intelligence Toolbox for Earth Observation (AiTLAS), both of which are actively used in the field of aerial archaeology. RVT is indispensable for processing input data (for training and inference) by converting digital elevation models into ML-friendly visualisations, while AiTLAS provides access to the ML models. We have conducted a series of experiments with different visualisation methods and different ML architectures for object detection and semantic segmentation to find the optimal configurations for the software.
The tool and its use will be presented by a core member of the development team, Žiga Kokalj.