We’re excited to share some groundbreaking work our systems librarian Tim Graves has been doing in collaboration with Danny Millum from our Collections team and DISCUS, the Data Intensive Science Center on campus. His focus has been on leveraging the latest advancements in artificial intelligence (AI) to unlock the hidden treasures within our collections.

Watch his latest video – Using AI to explore collections at the University of Sussex

The Power of AI in Uncovering Hidden Information

Traditionally, cataloging has its limitations. Important information often remains buried, only discoverable through meticulous, manual search. However, AI offers us a way to delve into collections and unearth content that traditional methods might miss.

One of our goals is to develop tools that are versatile and adaptable, capable of analyzing any collection of materials. While our current focus is on a collections from Tricontinental, the techniques we’re refining will be broadly applicable.

AI-Assisted Research

Imagine you are a new researcher approaching the Tricontinental collection with no prior experience. How can AI assist you?

1. Keyword Analysis and Heat Maps: By feeding several issues of Tricontinental into an AI, we can generate a heat map showing the frequency of common terms. This provides a quick overview of prevalent topics, though the context is essential for deeper understanding.
2. Semantic Mapping: DISCUS is helping us create semantic maps that analyze words in context, matching them against established subject groupings. This approach promises to reveal the themes and subjects covered in the collection more comprehensively.
3. Overcoming Language Barriers: Many documents in collections like Tricontinental are in various languages. AI tools can translate text instantly, breaking down language barriers that would otherwise hinder research. For instance, using large language models, we can translate Spanish or even Arabic texts into English, making them accessible to non-speakers.
4. Visual Image Analysis: Tricontinental contains numerous images and posters. AI can identify and describe these images, providing context and additional information. For example, we can use AI to recognize figures like Che Guevara in posters and describe the content of these images, enriching our understanding of the visual materials.

Enhancing Discoverability

One of the most exciting potentials of AI is its ability to enhance the discoverability of collections. Traditional cataloging picks out key subject terms, but AI allows us to probe documents at the word level.

Tim created a demo chatbot that can search through the content of several issues of Tricontinental. This chatbot responds to natural language queries, offering a more intuitive and human-like interaction with the collection. For example, you can ask about famous individuals mentioned in the documents or the subjects related to a specific figure like Castro. The chatbot provides succinct, accurate answers based on the content it has analyzed.

Future Aspirations

What can we aspire to use AI to do:

• Discover detailed semantics and topics within collections.
• Overcome language barriers for multilingual materials.
• Catalog and describe images accurately.
• Implement a chatbot interface for a more conversational exploration of collections.

This AI-driven approach transforms how we interact with and understand our collections, making research more accessible and insightful. We’re excited about the possibilities and look forward to sharing more updates as our librarians continue our journey into the future of library sciences.

For more information and a deeper dive into our work, check out the video here.

Stay tuned for more updates!

Tim Graves is the Systems Librarian at the University of Sussex Library, spearheading innovative projects to enhance research and accessibility through advanced technologies.