30 November 2025 to 3 December 2025
Century City Conference Centre
Africa/Johannesburg timezone
PLEASE NOTE: Registrations Have Closed! Contact chpc@csir.co.za for further queries.

Intelligent Metadata: Harnessing AI for Discovery, Visibility, and Preservation in the Digital Library

Not scheduled
20m
BICC.G-D2 - D2 Tsitsikamma (Boardwalk Convention Centre)

BICC.G-D2 - D2 Tsitsikamma

Boardwalk Convention Centre

120
Talk DIRISA DIRISA

Speaker

Mr van der Walt Isak

Description

The University of Pretoria’s Department of Library Services (DLS) has developed an inhouse
Artificial Intelligence (AI) Toolkit to enhance metadata creation and enrich digital
collections. Aligned with the theme “From Data to Decisions: Leveraging Cyber
Infrastructure,” this initiative uses AI to strengthen discovery, visibility, and preservation
across institutional repositories and special collections.
The toolkit integrates multiple AI capabilities—image and video analysis, scene
detection, emotion recognition, transcription, keyword generation, Sustainable
Development Goals (SDG) classification, and object detection—to automate and
improve metadata quality. Through scene and object detection, audiovisual materials are
segmented and described according to identifiable elements such as people, settings, and
activities. Emotion recognition provides contextual insight into affective dimensions in
recordings, while speech-to-text transcription and keyword extraction enable automated
summaries and thematic indexing, reducing manual cataloguing effort and improving search
accuracy.
The inclusion of SDG classification models, trained on BERT-based architectures, links
research and digital outputs to global sustainability goals, supporting institutional reporting
and impact measurement. The toolkit’s modular architecture ensures integration with
existing library systems and adherence to privacy and ethical standards while promoting
scalability across various workflows, from digitisation to research dissemination.
By embedding AI into metadata workflows, the DLS transforms static data into actionable
insight, aligning digital preservation with decision-making and strategic research visibility.
The outcomes include:
Page 2 of 2
• Improved discoverability and interoperability of digital assets.
• Increased efficiency in metadata generation.
• Enhanced contextual understanding of multimedia collections.
• Strengthened alignment with global open scholarship and sustainability
frameworks.
This initiative demonstrates how AI-driven metadata enrichment can empower libraries to
move beyond traditional curation, positioning them as key actors in data-driven innovation
and digital scholarship within the modern academic ecosystem.

Institute University of Pretoria
Presenting Author Mr Isak van der Walt

Primary author

Mr van der Walt Isak

Presentation Materials