Loading…
Tuesday September 24, 2024 11:00am - 12:30pm CEST

This session consists of 4 presentations and a joint Q&A with the presenters. The session contains:

➺ Johan Oomen, Rasa Bocyte - Insights from Practitioner Dialogues: Implementing AI Solutions Across Diverse Organisational Contexts (Short presentation)

➺ Isabel Schellnack-Kelly, Mashilo Modiba - Developing Smart Archives in Society 5.0: Leveraging Artificial Intelligence for Managing Audiovisual Archives in Africa (Long presentation)

➺ Alexander Wolff, Kim Voss - Is AI good enough? The struggles of automating a historical media archive. (Long presentation)

➺ Desiree Alexander, Chloe McLaren - AI for AV: Leveraging Artifical Intelligence for accessibility and discoverability of our audiovisual collections (Short presentation)


**Abstracts:**


➺ Insights from Practitioner Dialogues: Implementing AI Solutions Across Diverse Organisational Contexts
--
Johan Oomen, Rasa Bocyte (Short presentation)
--
This presentation explores research findings regarding the integration of AI solutions across various organisational contexts. The transformative potential of applying artificial intelligence in the context of audiovisual archiving has been demonstrated across a spectrum of use cases related to search & exploration, preservation, artistic expression and big-data analysis. Policy level interventions, such as the AI Act and the Ethical Guidelines for Trustworthy AI, also help to shape institutional strategies toward the responsible utilisation of AI technologies.
However, earlier research showed that integrating AI in the audiovisual domain still comes with a range of challenges. Key among these challenges are inquiries such as: (1) the strategic choice between off-the-shelf and bespoke solutions; (2) aligning the requirements of AI technologies with dimensions of existing (legacy) systems; (3) aligning public values & ethical considerations in procurement processes; (4) the scalability and-long term viability of solutions; (5) the imperative of fostering AI literacy to facilitate decision-making.
Given the specificity of the audiovisual domain (f.i. encompassing the time-based nature and scale of collections, adherence to specific standards, legacy data, and legacy systems) it is essential to develop and exchange good practices. With this in mind, the AI4Media Network of Excellence orchestrated dialogues with diverse practitioners from the media sector to capture their experiences. During this presentation, we will discuss these insights to provide the IASA community with practical guidance on sustainable and responsible integration of AI solutions into organisational workflows. Along with the formulation of organisational policies to guide the selection and utilisation of AI technologies.

➺ Developing Smart Archives in Society 5.0: Leveraging Artificial Intelligence for Managing Audiovisual Archives in Africa
--
Isabel Schellnack-Kelly, Mashilo Modiba (Long presentation)
--
In the age of Society 5.0, marked by the profound integration of digital technologies and human society, the preservation and management of audiovisual archives in Africa face substantial challenges. The exponential growth of digital content, combined with the continent's diverse cultural landscape, presents complex obstacles to conventional archival practices (Smith, 2020). Without the effective implementation of Artificial Intelligence (AI) technologies, crucial aspects such as curation, accessibility and sustainable management of audiovisual archives in Africa remain hindered. Thus, the overarching problem under scrutiny pertains to how the establishment of intelligent archives, facilitated by AI, can offer innovative solutions for preserving and managing audiovisual archives in Africa, within the overarching framework of Society 5.0. This approach has the potential to significantly contribute to safeguarding Africa's rich cultural heritage and advancing research, education and cultural exchange. The purpose of this paper will be to focus on the challenges encountered by selected African countries in managing their audiovisual collections and what can be done to facilitate AI. It was result in the formulation of a framework for implementing smart archives using AI in African contexts, emphasizing practical strategies and best practices.

➺ Is AI good enough? The struggles of automating a historical media archive.
--
Alexander Wolff, Kim Voss (Long presentation)
--
In 2022, the German Broadcasting Archive founded a specialized department for automation tasks to reach the goal of having all its collections documented until 2034. Since then, the team has developed strategies to help reach this ambitious goal with a lot of creativity, that led to successful projects and tools in the long run.
With large language models and other AI-driven tools, we lately discovered a whole new world of opportunities for our archive. Still, the special character of our heritage from the German Democratic Republic and the early times of German radio broadcasting is challenging when it comes to applying AI tools, that are trained on current data from the web. Also, together with our metadata experts we discuss if it is possible to reach the high-quality metadata standards for our databases with AI tools.
We tackle these challenges with various approaches and have become experts in tasks like prompt engineering and automatic transcription. Our tools increase the effectiveness in documentation and change the way information specialists work by supporting them with AI and non-AI-based data mining.
In our talk, we will present the struggles we overcame in applying AI and automation of historical media heritage and the tools we created and are still creating to automate the archive's daily work processes. We will also point out what this means for the work culture of a historical archive when old work routines are interrupted by new technologies.

➺ AI for AV: Leveraging Artifical Intelligence for accessibility and discoverability of our audiovisual collections
--
Desiree Alexander, Chloe McLaren (Short presentation)
--
Automated speech recognition (ASR) has evolved significantly in recent years, offering organizations stewarding AV materials unprecedented opportunities to leverage these tools to increase the accessibility and discoverability of their digital collections.
In this session, we will discuss our experience at Cornell University Libraries in evaluating and implementing ASR tools into our preservation workflows, with a specific focus on Whisper AI, an open source, command-line utility developed by OpenAI. We hope attendees will gain insight into the strengths and weaknesses of leveraging ASR tools, the challenges and opportunities presented by the technology, and practical guidance on how to develop or expand AV accessibility-related projects.

Moderators Speakers
avatar for Johan Oomen

Johan Oomen

Manager Research & Heritage Services, Netherlands Institute for Sound & Vision
As Head of Research and Heritage Services at the Netherlands Institute for Sound & Vision, Johan Oomen spearheads efforts to provide access to digital heritage. Additionally, he contributes as a researcher at the User-Centric Data Science group of VU University Amsterdam. Next to... Read More →
avatar for Rasa Bocyte

Rasa Bocyte

Researcher, Netherlands Institute for Sound & Vision
Rasa Bocyte (she/her) is a Senior Advisor for Research Collaborations at the Netherlands Institute for Sound & Vision. She is passionate about forging cross-sectoral collaborations underpinned by creative and critical approaches. In her current role, she leads the development and... Read More →
avatar for Isabel Schellnack-Kelly

Isabel Schellnack-Kelly

Professor - Archives and Records Management, Department of Information Science, Unisa, University of South Africa
Prof ISABEL SCHELLNACK-KELLY is a professor in Unisa’s Department of Information Science. She teaches modules related to archives, records management, audiovisual archives and supervises Masters and Doctoral students. She has vast experience in archives and records management having... Read More →
MM

Mashilo Modiba

Dr Mashilo Modiba is a senior lecturer in the Department of Information Science at UNISA. His PhD research focused on “Utilising artificial intelligence technology for the management of records the Council for Scientific and Industrial Research in South Africa”. His research interests... Read More →
avatar for Alexander Wolff

Alexander Wolff

Data Scientist, Deutsches Rundfunkarchiv (German Broadcasting Archive)
Alexander Wolff is a founding member of the Automation team at the German Broadcasting Archive. As such, he has established the necessary foundations of automation processes and worked on various projects in this field. He is self-educated in Python and AI and started this passion... Read More →
avatar for Kim Voss

Kim Voss

Data Scientist, German Broadcasting Archive
Kim Voss is an information specialist and product owner of automation projects at the German Broadcasting Archive. In her current projects, she is working on integration of automation and AI in documentary workflows. With her expertise at the intersection of media documentation, data... Read More →
avatar for Desiree Alexander

Desiree Alexander

AV Collections Coordinator, Cornell University
Desi Alexander is the Audiovisual Collections Coordinator at Cornell University where she manages the AV Preservation Lab, working to identify, prioritize, and digitize the library's at-risk and obsolete media formats. In this role she works with a diverse range of content and stakeholders... Read More →
CM

Chloe McLaren

Metadata Librarian, Cornell University Library
Chloe McLaren is the Metadata Projects Librarian at Cornell University Library. She currently works broadly in the institutional repository and manages the library’s streaming video platform and is engaged with digitization workflows. Previously, she was the Technical Manager at... Read More →
Tuesday September 24, 2024 11:00am - 12:30pm CEST
Classroom 1

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Share Modal

Share this link via

Or copy link