Blame the Algorithm?!

A FIAT/IFTA Travel Grant 2023 Blog!

Mutanu Kyany’a from African Digital Heritage on the left, posing for a ‘this is for the archives’ photo with Eppo van Nipsen and Julia Vytolpil from the cool gang at Sound and Vision.

By Mutanu Kyany’a (African Digital Heritage)

Picture this: Sharing a meal with media archiving enthusiasts from around the world in the heart of Locarno, Switzerland, a place known for its cinematic glory. That’s just one of the many unforgettable moments I took home from the 2023 FIAT/IFTA conference held at the PalaCinema. It felt like the perfect homage to the world of film; gathering professionals dedicated to preserving our rich audiovisual heritage. As a newbie at this conference, it was an honour to connect with seasoned members of the FIAT/IFTA community, some of whom have been in the game for much longer.

Mutanu Kyany’a from African Digital Heritage on the left, posing for a ‘this is for the archives’ photo with Eppo van Nipsen and Julia Vytolpil from the cool gang at Sound and Vision.

Mutanu Kyany’a from African Digital Heritage on the left, posing for a ‘this is for the archives’ photo with Eppo van Nispen and Julia Vytopil from the cool gang at Sound and Vision.

The event was centred around the fusion of AI with the world of media archives. It’s amazing to see how this technology is making waves in the realm of audiovisual archiving. The discussions covered a wide array of innovative ways AI is being woven into archival practices, each shedding light on the possibilities and benefits.

AI for Content Cataloguing

One standout topic was the role of artificial intelligence in cataloguing and organizing vast archives. AI is revolutionizing the way we handle transcription, labelling, and categorization of audiovisual materials. The SAVA team shared their journey of using AI to streamline cataloguing by identifying performers in videos through OCR and building thematic databases. Similarly, RTVE highlighted their venture into AI for automating cataloguing in their television archive. The result? Better content descriptions, improved accessibility, and a surge in the use of archival material.

AI for Metadata Enhancement

AI isn’t just about automation; it’s a game-changer for enhancing metadata associated with archival content. From identifying and correcting offensive terms to adding relevant keywords, AI is elevating the quality of metadata. The DE-BIAS project, funded by the EU, demonstrated how AI can automatically detect and correct offensive terms in cultural heritage metadata. What makes it even more impressive is that they’re collaborating closely with the communities linked to the subjects in the archive, ensuring a holistic and inclusive approach.

AI for Content Retrieval

Imagine having AI-powered search algorithms and similarity matching to help you navigate vast archives. It’s like a treasure hunt made easy. The Netherlands Institute for Sound & Vision showcased their implementation of AI-based similarity searches. This enables users to find matching scenes by providing sample images, which significantly improves content retrieval. What’s heart-warming is the real-life stories that emerge from these finds; someone even recognized themselves in an old photo, unlocking new narratives and memories.

AI for Preservation and Migration

AI isn’t just about organizing and retrieving; it’s also a superhero in preserving digital media and migrating content to ensure long-term accessibility. The European Broadcasting Union introduced EBUCorePlus, an open-source tool that uses AI and semantics to enhance media asset preservation. It covers the entire media value chain, seamlessly integrating various aspects like planning, production, distribution, and archiving. What’s more, it can integrate external information sources, thanks to its knowledge graphs and linked open data sources.

Sure, the abundance of AI projects may seem overwhelming, especially for those of us who are yet to digitize our archives. But as I delved into these ideas, I couldn’t help but feel a sense of hope. It’s clear that now is the perfect time for those of us who haven’t gone digital with our media archives to actually do it! The experimentation has already been done, and the best practices are laid out. We can digitize carefully curated archives that reflect our own realities and identities, with the invaluable guidance of pioneers in the field.

But here’s the catch: How do we ensure that our digitized archives stay relevant and adapt to future technologies? How do we prevent ourselves from falling into the trap of blaming r=the algorithm and consequent AI tools when they don’t deliver the ‘right’ results?

It starts with solid cataloguing practices. We need cataloging frameworks that accurately represent our societies and knowledge systems; making data easily searchable, accessible, and adaptable for repurposing. Collaboration with our communities is vital to capture accurate metadata and to, in the words of Asli Ozgen, help us activate the archive, and in turn they can be activated by the archive.

Blame the Algorithm?!

We also need the expertise of tech wizards who can create tools aligned with our unique contexts and realities. And let’s not forget the importance of advocacy from civil society and lawmakers, who will help us with relevant traditional knowledge laws that will protect our data.

I’m optimistic that we can not only get this done but get it right. The chance to present our histories in engaging, immersive, and educational formats is now! The opportunity is ripe to change the perception of our archives from mundane to captivating, allowing our communities to see them in a whole new light.

As one presenter wisely put it, “An archive is a time travel machine to self-discovery.” So as we build these time travel machines, let’s do it right.

See you at the next FIAT/IFTA conference in Bucharest! 😊