A GLAM collection manager#
The following scenario is based on the Head of Digital Innovation at a museum or gallery whose primary role is to secure funding and develop a strategic vision, not to write code. This individual needs to articulate the value of NER to a non-technical audience, outline a clear project plan, and demonstrate how this work directly supports the institution’s mission.
The museum is launching a new exhibit called “Echoes of History” to create a dynamic, interconnected experience for visitors. This exhibit will use Named Entity Recognition (NER) tools to reveal hidden relationships between the museum’s own collection and artifacts, events, and people in the wider world. The goal is to showcase how these tools can identify, clarify, and link entities to a broader global narrative, enriching the visitor’s understanding of each piece.
The Challenge: Connecting the Collection to the World#
The museum’s curator, Alex, wants to move beyond showcasing the collection as a series of isolated objects. The true stories are often found in the connections—how a painting in the museum was influenced by a work in a different country, or how a sculptor was inspired by a historical event that isn’t represented in the museum’s collection. The challenge is to automatically and intelligently create these connections to entities outside the museum.
The “What”: A New Exhibit Experience#
Alex envisions an exhibit where visitors can choose an artwork and instantly see its global context. This isn’t just about showing a painting’s title and artist; it’s about revealing a web of connections to the outside world. Recognition & Disambiguation: The NER tool first scans the description of a painting, such as “The Night Watch” by Rembrandt. It recognises key entities like “Rembrandt,” “17th century,” and “Amsterdam.” The tool then disambiguates these entities, confirming that “Rembrandt” refers to the Dutch painter and “Amsterdam” to the city where he worked. Linking to External Exhibits & Knowledge Bases: This is the core of the new exhibit. Instead of just linking to other works in the museum, the tool automatically links to entities outside the museum’s collection. For example, “Rembrandt” is linked to his biography in Wikidata and to the Rijksmuseum, where many of his other works are held. “17th century Amsterdam” is linked to external historical records about the city’s golden age, providing context for the painting’s creation. The tool might also identify and link to less obvious connections, such as a historical event like the Dutch-Portuguese War, which could have influenced the subjects in the painting. The “Echoes” Experience: Visitors interact with a digital kiosk next to the artwork. The screen displays the artwork’s description along with a dynamic visual map of its “echoes.” A line might connect “The Night Watch” to the Rijksmuseum, with a small thumbnail of that museum. Another line might connect it to the Portuguese settlement of Macau, an important trading post for the Dutch, highlighting a less-known aspect of the historical context. The visitor can then click on any of these external links to explore them further.
The “Why”: The Value Proposition#
This approach provides a powerful value proposition for both the museum and its visitors: For Visitors: It transforms a static exhibit into a global learning journey. It allows them to understand a single piece of art not in a vacuum, but as part of a rich, interconnected tapestry of human history and creativity. It makes the museum’s collection feel more relevant and alive. For Funders & Stakeholders: This demonstrates the museum’s use of cutting-edge technology to achieve a unique educational and curatorial goal. It positions the museum as a leader in digital innovation and global collaboration, as it leverages external knowledge bases and information from other institutions to enhance its own. It’s a scalable solution that can be applied to the entire collection, offering a significant return on investment. This approach shifts the focus from a self-contained collection to one that is part of a larger, global conversation.
The Challenge of Connecting the Collection#
Connecting a museum’s collection to the broader world using NER is a deceptively difficult problem, which is why it hasn’t been widely implemented yet. The main challenge lies in the unstructured and often obscure nature of historical data. Unlike modern web content, museum archives contain a mix of formal catalogues, handwritten notes, and digitised letters, often using inconsistent terminology. A place name like “Florence” could be a city, a person’s last name, or a street, and the context needed for disambiguation is frequently missing. Furthermore, historical figures, places, and events may not have clear, standardised entries in modern knowledge bases like Wikidata. This creates “orphan” entities that the linking tool can’t match, or it produces false positives. The process is also computationally intensive and requires continuous human oversight to train the models on specialised vocabulary (like art movements, obscure artists, or technical art terms) and to correct the inevitable mistakes. In essence, it’s not just a technological challenge, but a deeply complex curatorial one that requires a blend of machine learning expertise and historical knowledge.
The AI recipes#
The AI recipes we are developing can help by providing a standardised, modular framework for addressing the complex challenges of applying named entity recognition and disambiguation to cultural heritage data. Instead of building a bespoke solution from scratch, museums can follow a recipe that guides them through each step, from data preprocessing and annotation to model training and evaluation. This approach democratises the technology, making it accessible to institutions with limited technical staff by offering pre-trained models fine-tuned for the unique vocabulary of art history and museum archives. The cookbooks also include methods for evaluating and validating the results against a “gold standard,” ensuring the connections made are accurate and reliable, thus building trust in the technology. This not only accelerates the research and development process but also ensures that the final product is both effective and aligned with the museum’s curatorial and educational goals.