Creating an Interactive Digital Collection with an AI Chatbot

Custom AIs can bring digital collections to life

Posted on March 17, 2025 in Blog Posts

Recently, my student assistant, Tharun Viswanathan, and I became intrigued by the potential of using chatbots with digital collections. What started as a small inquiry quickly evolved into a significant project: a combined digital and physical exhibit based on our collection of Cleveland Bystander publications. The result was a novel way of interacting with digital collections, a deeper understanding of the scope of such projects, and an opportunity to make AI-driven initiatives more accessible to cultural heritage institutions with large, text-based digital collections.

Creating a chatbot from a collection

Formerly known as Cleveland Town Topics and Country Club News before merging into one publication in the late 1920s, the Cleveland Bystander was “a monthly publication which covered social and cultural news from Cleveland’s high-society residents in the late 19th and early 20th centuries,” as described on Digital Case, Case Western Reserve University’s digital library, where this digitized collection is openly available to view and download. This collection comprises 88 digitized volumes, totaling 4.3TB for just the access copies available on Digital Case. 

Digitizing this collection was a massive milestone for the library, and it remains a large and valuable source of textual data. Having this collection fully digitized and available came at the perfect time, with libraries and cultural heritage institutions looking for ways to integrate AI into their field—a very timely topic at Case Western Reserve University Libraries. From serving as a pilot project to advocate for the library’s need for high-performance computing (HPC) resources to showing how large language model agents can do retrieval augmented generation (RAG) from different data models (such as a CSV, XML, or JSON file), the Cleveland Bystander Collection is truly becoming a first in many areas for the library as we continue to investigate how artificial intelligence can be implemented. 

This is further exemplified in the creation of the Bystander Chatbot, which leverages the textual data from the historical newspaper as the input for this large language model and allows users to interact with this large amount of text without having to scroll through the hundreds of thousands of pages manually. The Bystander Chatbot is an innovative, AI-driven chatbot designed to bring to life the rich social and cultural history of this regional newspaper. This chatbot engages audiences through enhanced access, improved discoverability, and an interactive experience to dive deep into Cleveland’s history. By combining this rich digitized collection with cutting-edge technology, the chatbot represents a groundbreaking step forward in enhancing the way we access, discover, and interact with digital collections. This blend of historical depth and modern innovation offers a unique window into Cleveland’s illustrious past, making history more vibrant and approachable for all. 

However, blending historical depth and innovation can be challenging, time-consuming, and a bit of a whirlwind. We quickly realized we had underestimated the project’s scope. While exporting the textual data from the digital library was straightforward, and the initial code and chatbot setup progressed rapidly, we soon encountered issues with the user interface and hosting. After evaluating multiple solutions, we decided to host the chatbot independently through Amazon Web Services (AWS), intending to make it available for a limited time (until the end of April) due to financial costs and sustainability concerns. Initially, we aimed for a flashy user interface with impressive UX features, but we ultimately opted for a practical, user-friendly text message interface. This familiar and relatable approach was more manageable and ideal for increasing users’ AI literacy.

Although we decided not to keep the chatbot as a permanent resource, we recognized the value of making Tharun’s code openly available for institutions and others to use with their own digitized collections. While the hosting details are not included, this leaves room for further opportunities, improvements, and customization. Our primary goal was to explore the feasibility of integrating digitized collections into a chatbot to enhance user interaction with text-focused collections. Thankfully, it was possible, and we discovered how AI could be implemented in various ways within the library.

---

🔥 Stay up-to-date with LibTech Insights by signing up for our free newsletter. Just one weekly email with our new blog posts, top tech news stories, and other bonus content. Check out some posts from our archive:

• Balancing User Needs and Technology during Migration Planning
• How Archetypes Can Help Focus Your Library’s Website on User Needs
• Using AI to Create Better AI Prompts

🦋 Choice is now on Bluesky. Say hi to us over there!

---

Fostering collaboration in the library

The technical aspects of this project took a large amount of time, but the Bystander Chatbot brought to light a great opportunity to collaborate and increase engagement with digital collections across the library. One of my main goals was to showcase Tharun’s hard work as a student assistant throughout this long process, from an initial idea to a fully fledged chatbot. It was essential to promote this chatbot and have it be known across the library and campus that this work was being done, especially given the timeliness of AI. I was certainly not quiet when it came to telling others about Tharun’s work and how exciting it was to see progress. 

This prompted the Exhibits Librarian and Campus and Community Outreach and Engagement Manager to reach out and inquire about making this project into an exhibit in the library, offering more ways to inform library users about the collection, the chatbot, and the amazing work that the library’s student assistants are conducting. We determined that while the chatbot (and corresponding textual data in the collection) was clearly the superstar, we did not want users to ignore the physical aspect of this collection, especially because, without the physical copies of this historical newspaper, we would not have digital surrogates or anything to digitally harvest the data. With the idea in mind of honoring both the physical and digital aspects of this project, we made it a point to also showcase the amazing cover art across these digitized volumes. The bright, beautiful cover designs and the tremendous research value of this collection were a perfect match to put on display and promote the digital collection and chatbot. 

We also enlisted the expertise of the Team Lead for University Archives and Special Collections, whose extensive knowledge of the collection was invaluable. They crafted a captivating description of the Cleveland Bystander, providing users with rich contextual information about the foundation for the chatbot. To build anticipation, we displayed the stunning cover art and detailed descriptions of the collection early on, offering a preview before the grand unveiling of the exhibit, “Echoes of Elegance: Rediscovering Cleveland’s High Society Through AI.” By broadening the focus beyond the AI elements to include the physical artifacts, their historical context, and accessibility, this exhibit allowed us to highlight the collection’s value throughout its entire information lifecycle. This holistic approach underscored the necessity of one final collaboration to bring the exhibit to its full potential.

Highlighting the physical collection through its beautiful cover art and providing background information on its historical context offered users valuable insights into interacting with the chatbot. Not all users may want to use the chatbot; however, some may view this collection as a bridge to their own research methodologies and practices. This exhibit presented a perfect opportunity to involve the library’s Digital Scholarship team. This collaboration not only expanded user interaction with the collection but also promoted the DS Team’s services and resources related to this type of data. The DS Team compiled a list showing various research applications for the collection beyond the chatbot, such as entity recognition, network analysis, sentiment analysis, and more. Including these details enriched the exhibit, making it multifaceted and appealing to a broader audience.

Takeaways

Overall, this project has been a sort of revelation, showcasing how we can elevate our creativity through AI. The unexpected challenges we faced only underscored the importance of continuous testing and interaction with these innovative resources. Each obstacle became a stepping stone, reinforcing our commitment to exploring how AI can revolutionize our services and enrich our users’ experiences. Even though our expectations were larger when beginning this project, the journey of this chatbot not only broadened our horizons but also ignited a passion for discovering the limitless possibilities that AI holds for the future of digital collections and how users interact with these collections in a continually evolving technological landscape. 

Related Posts