LadeCA
Informatics conveys art

dialog

I would like to use these 5 minutes to present my new project, namely the development of an AI-driven dialog system for communicating art collections.

Visual art collections are more than just sets of images. Behind a collection or an exhibition are the curator’s concept and the specific characteristics of the collection. It is obvious that in the case of collections with thousands of images it is not possible to present the images side-by-side or in a slide show. There is therefore a need for a system for communicating such collections with regard to the curator's concept and the specific characteristics of the collections.

The initial situation for the system is a collection of visual art and users with an interest in the collection. The first step is to create a thesaurus to describe the collection, that is, a structure of the collection that reflects the curator's concept and the specific characteristics of the collections. Taking into account the thesaurus and the reactions of a user, an AI then conducts a dialog with the user with the aim of transmitting the corresponding collection, taking into account the curator's concept, the specific characteristics of the collections, and the interests and experience of the user.

The dialog system consists of three parts, namely the dialog interface, the AI algorithm and the representation of the art collection. My project focuses on the dialog interface. As part of my project, I will implement and compare several different options. The aim of the AI part is to convey the curator's concept and the specific characteristics of the collections, but also to take into account the user's experience, knowledge and interest. One important function of the AI will be that the user can select any value on a scale ranging from a pure query system to a pure recommendation system. The part of creating a thesaurus for a given art collection has already been completed. With LadeCA it is possible to build up a controlled vocabulary that is expanded into a thesaurus representing the curator's concept and the specific characteristics of the collection. This part has already been described in detail and published in several articles.

The entire system can also be seen as an experimental kit with which parts of the system can be newly developed and be compared with existing solutions.

Pflüger, H.: A method for describing, analyzing, and exploring visual art collections (2024).
Link: https://dx.doi.org/10.18419/opus-13165

ladeca

Creating an initial controlled vocabulary

ladeca

Optimizing the granularity of a controlled vocabulary

ladeca

Exploring and structuring a controlled vocabulary

Pflüger, H.; Viehhauser, G. & Tumanov, R.: An interface to view collections of visual
art. In: International Journal of Digital Humanities (2023).
Link: https://doi.org/10.1007/s42803-022-00061-8

ladeca

Introduction video

Marie Herbert: Computergestützte Exploration und Analyse von van Goghs Ölmalerei:
eine Betrachtung und Charakterisierung seiner Arbeitsweise anhand von Bildausschnitten.
Masterarbeit (2022).
Link: https://dx.doi.org/10.18419/opus-13165

Pflüger, Hermann: A language to analyze, describe, and explore collections of visual art.
In: Visual Computing for Industry, Biomedicine, and ArtVisual Computing for Industry,
Biomedicine, and Art
(2021).
Link: https://doi.org/10.1186%2Fs42492-021-00071-3