Multidisciplinary Research Proposal Enhancements through Large Language Models
By: James Creel, Daniel Xiao & Ethel Mejia
Following a TAMU Libraries directive to prepare for disruptions posed by emerging Generative AI technologies, the Office of Scholarly Communications developed a plan to use Scholars@TAMU profiles—supplemented with data on grant funding and other initiatives—as inputs to a Large Language Model (LLM) for identifying serendipitous collaboration opportunities.
Despite growing expectations for forming crosscutting collaborations, research development units often struggle to proactively form multidisciplinary teams. Research Information Management Systems (RIMS) contain rich, structured data on faculty expertise, but are rarely leveraged for systematic researcher matchmaking. Our LLM-supported workflow addresses this gap by integrating RIMS data into a proactive and transparent mechanism for surfacing interdisciplinary connections.
The workflow proceeds in three stages: (1) extracting relevant Scholars@TAMU profiles based on keywords associated with a research opportunity; (2) submitting profile data and contextual information to an LLM using a tailored prompt to generate potential collaborations and recommended actions; and (3) curating these outputs for possible dissemination to researchers. The workflow remains experimental, and key questions remain regarding its robustness, integrity, and value for scholarly communications.
We are designing experimental scenarios to evaluate the utility and reliability of this workflow. This includes defining profile sets and query variants, identifying suitable LLM platforms, and accounting for the probabilistic nature of LLM outputs through redundant runs. We aim to validate the methodology in partnership with researchers willing to engage with the generated recommendations. In this presentation, we will share our methodology, preliminary results, and any code or related resources.
Building A Removable Media Workstation For Accessing Floppy Disks and More
By: Rose Goldey
Like most archives and special collections, Texas State University hold born-digital materials stored on legacy removable media such as CD and DVDs, external hard drives, and floppy disks. However, the institution has historically lacked experience or workflows on how to recover, reformat, and access their contents.
This lightning talk describes the development of an air-gapped removable media workstation from initial conversations in early 2024 to a fully-working system in late 2025. Topics included are; configuring a standalone system outside the institutional network, establishing basic handling and documentation practices, and evaluating tools for different media types. The talk highlights early work with optical media, including CD and DVD ripping, as well as the challenges encountered when attempting to recover data from floppy disks.
One of the biggest pushes for this removable media workstation was the ability to access floppy disks. Initial work with USB-floppy drives allowed access to the disks, but showed errors and incomplete data. The solution was the use of a Greaseweazle floppy disk controller to enable flux-level imaging. Rather than presenting this tool as a universal solution, the talk discusses the ongoing process of refining our born-digital workflows.
The presentation concludes with discussion of future directions, including expanded digital processing workflows, continued learning around floppy disk repair and restoration, and the ongoing development of sustainable, defensible practices for born-digital archival materials.
Lab Band Digital Debut: Digitizing UNT’s Jazz Sheet Music Library
By: Steven Sellers & Marcia McIntosh
Since the inception of the University of North Texas’ Jazz Studies program in the late 1940’s, students and alumni have contributed to its rich musical history in the form of original student arrangements played by the world famous and Grammy nominated “Lab Bands”. Over 75 years later, the collection of arrangements continues to grow, currently reaching a staggering eight thousand charts. After years of aspirations to digitize this collection, in early 2025 an opportunity has presented itself by way of collaboration between the College of Music, the UNT Music and Digital Libraries, and the founders of the Sherman Jazz Museum. While the project is still underway, this lightning talk will discuss the process of the first few phases of the project, with particular focus aspects like scope, collaboration, project management tools, and creative solutions for digitization challenges encountered while working on such a unique collection.
Bridging the Feedback Gap: Institutional Repository Training for Non-Institutional Repository Staff
By: Whitney Johnson-Freeman & Viktoriia Savchenko
At the University of North Texas Libraries, the institutional repository (IR) currently uses a mediated deposit that relies on submitters emailing their file along with some basic metadata. This process is meant to be a low barrier to contributing to the IR, and it enables the IR support team to easily communicate with individuals as they submit their work. However, it also means that there is some variation in the submission process, and this variation creates some challenges for the IR support team that slows their workflows down. Updates needed to be made to submission guidelines and outreach material, but the IR support team wanted to do it in an informed way. The IR support team decided to create an interactive workshop where library staff and student employees could get the opportunity to learn about the IR, get firsthand experience in digital preservation, and create their own metadata. Workshop attendees are then asked to share feedback with their new, more informed perspective, and the IR support team can use their feedback to improve their guidelines and outreach material. The goal is to have perspectives outside of the IR help bridge the gap between what the IR needs and how the support team communicates these needs. This presentation provides an overview of the workshop, the development process, and the outcomes so far.
Prerequisite Knowledge: Intermediate: Session is designed for attendees who have a basic understanding of the topic and some prior experience. It will build on core concepts and introduce more complex applications.
