Harnessing AI for Museum Collections at the National Museum of the Royal Navy

The National Museum of the Royal Navy explored two collection challenges with University of Southampton students: extracting useful search terms from written descriptions, and classifying photographs of warships. The experiments brought together collection expertise and AI methods to investigate how growing records could become easier to find and use.

The Role of AI in Museums

Museums house an enormous amount of information, from physical objects to archives and multimedia collections.

As custodians of knowledge, museums face the challenge of making their collections accessible to a global audience, both physically and digitally.

The National Museum of the Royal Navy has long embraced this responsibility, managing extensive collections that span archives, libraries, photographic, film, and sound records.

However, with the growing complexity of managing these datasets, traditional methods of cataloging and making collections accessible are becoming increasingly time-consuming.

Amy Adams, Collections Information and Access Manager at the National Museum of the Royal Navy, has witnessed this shift over her decade in the role.

She explains that “the amount of data we generate, and expect curators to manage, has increased drastically compared to even 5-20 years ago, when records were limited to a few lines of description in a card catalog.” Now, the expectation is to generate detailed information that can be utilised in numerous ways, from research to public education.

This is where AI comes in. For museums, AI offers the possibility of speeding up the process of data management while increasing accuracy and accessibility. The National Museum of the Royal Navy’s work with AI focuses on two primary areas: entity recognition and image classification.

Extracting search terms from descriptions

In the first of its AI projects, the museum sought to use entity recognition to streamline the management of its textual data. Traditionally, curators are tasked with writing descriptions for objects in the museum’s collection.

These descriptions are then used to generate metadata, or keywords, which make the collections searchable both internally and online. It’s a labor-intensive process that requires curators to anticipate which keywords users might search for—an inherently subjective task.

The museum worked with master’s students from the University of Southampton to explore how AI might assist with this. “We wanted to take the lovely descriptions our curators write and see whether AI could run those through and generate the relevant keywords,” Adams says.

For example, if a curator writes a description of a ship model, could AI extract keywords like “Seven Years War” or “Second World War,” and associate them with the object? Additionally, AI has the potential to think beyond the biases of curators, who might overlook more general search terms like “ship” in favor of technical terms.

The choice between general and specialist terms also raises the question of whose language appears in collection tags, explored in Jessica BrodeFrank’s work on crowdsourcing and AI.

Classifying the Wright & Logan photographs

Another project focused on the museum’s photographic collections, specifically the Wright & Logan collection. This archive of warship portraits, taken as ships entered and left Portsmouth Harbour, provided the perfect testing ground for image classification.

These images, largely standardised in their composition, were ideal for training AI to recognize different types of ships, such as destroyers or submarines. The museum hoped that AI could not only classify these images but also make the collection more accessible by identifying key features and generating relevant metadata automatically.

These projects explored ways to help curators manage growing volumes of collection data. Automating some time-consuming tasks could leave more time for interpreting and sharing history, while collection specialists would still need to assess the results.

The two experiments start with different material: descriptive text in one case and photographs in the other. They therefore ask different questions of the technology. Extracting a useful search term from a description is a separate task from recognising a ship type, and each still depends on collection expertise.

Human-AI Collaboration: The Key to Success

Despite the excitement around AI, Adams and her team are quick to emphasize that human expertise remains essential. AI might be able to generate keywords or recognize images, but it’s the curators who validate and guide these processes, ensuring that the technology produces meaningful results.

This human-AI collaboration is crucial in a heritage context, where domain expertise and nuanced understanding of historical objects are necessary to ensure the accuracy and integrity of collections data.

Dr Rodrigo Pacheco-Ruiz, a maritime archaeologist working with the museum, shared his perspective on how AI can complement the work of historians and archaeologists. “AI can help recognize ships in photographs, or keywords in descriptions, but it’s up to us to interpret what that information means and how it fits into the broader historical context.” He further emphasised that AI’s role is to assist, not replace, the human element in curatorial work.

The museum’s collaboration with the University of Southampton’s CORMSIS department is a prime example of how partnerships between the heritage and tech sectors can produce groundbreaking results. By working together, museum professionals and computer scientists are finding new ways to document and share history, while pushing the boundaries of what AI can achieve in the heritage space.

A Vision for the Future of Museums and AI

As these projects move forward, the National Museum of the Royal Navy is already thinking about the long-term implications of AI in the heritage sector. For Adams, the real excitement lies in the potential for AI to open up new ways of interacting with the museum’s collections.

Imagine asking a virtual assistant, like Amazon’s Alexa, about a specific ship, and receiving an answer powered by the museum’s collections data. This is no longer a distant dream but a realistic possibility as AI technologies continue to evolve.

Beyond public accessibility, AI also offers museums the chance to enhance their research capabilities. By pooling data from across institutions and combining it with external datasets, museums could uncover new insights about historical events, objects, and cultures. “We’re sitting on a lot of knowledge,” Adams says, “and AI could help us link it all together, making it easier to explore and understand.”

Of course, there are still challenges to overcome. As with any new technology, AI comes with its own set of limitations and learning curves. The museum’s initial work with AI has shown that while the technology is promising, it’s not a one-size-fits-all solution. Each project requires careful planning, testing, and adaptation to the specific needs of the collection.

Conclusion

The National Museum of the Royal Navy is proving that AI can be a powerful tool for heritage management, offering new ways to catalog, interpret, and share historical data.

By embracing technology, the museum is not only preserving the past but also paving the way for the future of museums and their collections.

As Amy Adams puts it, “This just feels like the beginning of the journey, rather than the end.” With the right partnerships and a continued focus on collaboration, AI will undoubtedly play an increasingly important role in the museum sector, helping institutions around the world unlock the full potential of their collections.