Metadatenerweiterung für Digitale Kulturgüter mit Large Language Models

DAVID WALSH; MARK HALL; DANIEL CAMPBELL; Paul Clough; Richard Palmer; Shehar Bano

doi:10.5281/zenodo.18702992

Metadatenerweiterung für Digitale Kulturgüter mit Large Language Models

Translated title of the contribution: Metadata Enrichment for Digital Cultural Assets with Large Language Models

DAVID WALSH
, MARK HALL^*
, DANIEL CAMPBELL
, Paul Clough
, Richard Palmer
, Shehar Bano

^*Corresponding author for this work

Computer Science

University of Sheffield
TPXImpact
Victoria and Albert Museum

Research output: Contribution to conference › Paper › peer-review

14 Downloads (Pure)

Abstract

The digitisation process has led to large digital archives of cultural heritage, but the metadata within these archives is often very limited. This limitation makes searching the archives more difficult and significantly restricts their usability. In this paper, we evaluate the usefulness of primarily local large language models for automatically identifying metadata relating to detailed aspects of digital cultural heritage objects. The results show that this approach is feasible, but that a final expert-driven selection remains essential.

Translated title of the contribution	Metadata Enrichment for Digital Cultural Assets with Large Language Models
Original language	German
Pages	570-571
Number of pages	2
DOIs	https://doi.org/10.5281/zenodo.18702992
Publication status	Published - 20 Feb 2026
Event	Digital Humanities in German-speaking countries - The University of Vienna, Vienna, Austria Duration: 23 Feb 2026 → 27 Mar 2026 Conference number: 2026 https://dhd2026.digitalhumanities.de/?page_id=18&lang=en

Conference

Conference	Digital Humanities in German-speaking countries
Abbreviated title	DHd2026
Country/Territory	Austria
City	Vienna
Period	23/02/26 → 27/03/26
Internet address	https://dhd2026.digitalhumanities.de/?page_id=18&lang=en

Keywords

DHd2026
digital cultural heritage
large language models
metadata
data recognition
content analysis
organising metadata

Access to Document

10.5281/zenodo.18702992Licence: CC BY

HALL_Mark_Metadatenerweiterung_f_r_Digitale_Kulturg_ter_mit_Final published version, 192 KBLicence: CC BY

SPOT: SPOT (Semantic Processing for Object Tagging): AI-Enriched Metadata for Cultural Heritage Collections
WALSH, D. (PI)
1/04/25 → 1/10/25
Project: Research

Cite this

@conference{88cf042041394f17a82f58488438fcf3,

title = "Metadatenerweiterung f{\"u}r Digitale Kulturg{\"u}ter mit Large Language Models",

abstract = "The digitisation process has led to large digital archives of cultural heritage, but the metadata within these archives is often very limited. This limitation makes searching the archives more difficult and significantly restricts their usability. In this paper, we evaluate the usefulness of primarily local large language models for automatically identifying metadata relating to detailed aspects of digital cultural heritage objects. The results show that this approach is feasible, but that a final expert-driven selection remains essential.",

keywords = "DHd2026, digital cultural heritage, large language models, metadata, data recognition, content analysis, organising metadata",

author = "DAVID WALSH and MARK HALL and DANIEL CAMPBELL and Paul Clough and Richard Palmer and Shehar Bano",

year = "2026",

month = feb,

day = "20",

doi = "10.5281/zenodo.18702992",

language = "German",

pages = "570--571",

note = "Digital Humanities in German-speaking countries ; Conference date: 23-02-2026 Through 27-03-2026",

url = "https://dhd2026.digitalhumanities.de/?page\_id=18\&lang=en",

}

TY - CONF

T1 - Metadatenerweiterung für Digitale Kulturgüter mit Large Language Models

AU - WALSH, DAVID

AU - HALL, MARK

AU - CAMPBELL, DANIEL

AU - Clough, Paul

AU - Palmer, Richard

AU - Bano, Shehar

N1 - Conference code: 2026

PY - 2026/2/20

Y1 - 2026/2/20

N2 - The digitisation process has led to large digital archives of cultural heritage, but the metadata within these archives is often very limited. This limitation makes searching the archives more difficult and significantly restricts their usability. In this paper, we evaluate the usefulness of primarily local large language models for automatically identifying metadata relating to detailed aspects of digital cultural heritage objects. The results show that this approach is feasible, but that a final expert-driven selection remains essential.

AB - The digitisation process has led to large digital archives of cultural heritage, but the metadata within these archives is often very limited. This limitation makes searching the archives more difficult and significantly restricts their usability. In this paper, we evaluate the usefulness of primarily local large language models for automatically identifying metadata relating to detailed aspects of digital cultural heritage objects. The results show that this approach is feasible, but that a final expert-driven selection remains essential.

KW - DHd2026

KW - digital cultural heritage

KW - large language models

KW - metadata

KW - data recognition

KW - content analysis

KW - organising metadata

U2 - 10.5281/zenodo.18702992

DO - 10.5281/zenodo.18702992

M3 - Paper

SP - 570

EP - 571

T2 - Digital Humanities in German-speaking countries

Y2 - 23 February 2026 through 27 March 2026

ER -

Metadatenerweiterung für Digitale Kulturgüter mit Large Language Models

Abstract

Conference

Keywords

Access to Document

Fingerprint

Projects

SPOT: SPOT (Semantic Processing for Object Tagging): AI-Enriched Metadata for Cultural Heritage Collections

Cite this