AI and archives – the human touch

The role of the archivist may be changing with the advent of AI-enabled services to enrich metadata, but there has to be human oversight to ensure the accuracy of data. That’s the view of leading library and media management expert, Kathey Battrick, talking to the Making the Media Podcast.

Kathey is the senior manager of library and media management with Asharq News, based in the United Arab Emirates, with a huge amount of experience in the archival field, dating back to the early days of creating digital libraries from film.

In the latest Making the Media Podcast, she describes how they are using AI services to enhance the metadata on material they are indexing and storing, and the benefits that brings in terms of discoverability for users.

But she emphasizes the importance of the work of the library team in this process. “We human-supervise the AI-generated content. So, the role of our archivists or news librarians as we call them here is to still create some manual cuts, but their role also is to supervise what the AI services are providing us with and correcting and fine tuning the metadata we get back.”

She adds, “It's also just constantly assessing and reviewing what you're getting back from those AI services and whether it's still relevant and still good enough. Are there any issues that need to be addressed, other improvements that could be made?”

Kathey also talks about the importance of recognizing different use cases that could be addressed when it comes to people searching to find relevant material.

“So, we will have colleagues working in the creative field, working maybe on promos or marketing content. They may be looking for more conceptual kind of shots, like a cityscape of Dubai or Riyadh in KSA, for example. So, it's going to be really important for them that they can find the landmark and cityscape that they might be looking for. Whereas we'll have our news production teams producing content about a specific event or breaking news story. And then, of course, we have our current affairs or documentary users that might be looking for content about a certain topic or place. So, it's a slightly more generic kind of search approach. And by utilizing AI, we can try to meet those different search requirements and sometimes use different platforms to access the content in an efficient and effective way.”

In the episode, she also talks about the need for multi-lingual metadata and her hope that more work will be done in the field of large language models (LLMs) which support Arabic to further enhance AI capabilities.

Find out more on the next episode of the Making the Media Podcast, available on all major podcast platforms on Friday, September 19.