Category: News

March 13, 202512:15 PM – 1:15 PM Speaker:William Mattingly, Cultural Heritage Data Scientist, Yale University Where:327 McVey Hall (Virtual Speaker)(Zoom link: https://uky.zoom.us/j/82467171189) Title:Evolution of Machine Learning Approaches in Human Rights Data Analysis Abstract:Over the past five years, our approach to named entity recognition and semantic search has undergone significant changes. In this tutorial, we will compare the methods used during the initial phase of Bitter Aloe with the techniques currently in use. Click here to see the complete list of speak

March 6, 202512:15 PM – 1:15 PM Speaker:Stephen Davis, Department of History, University of Kentucky Where:327 McVey Hall(Zoom link: https://uky.zoom.us/j/82467171189) Title:The Application of Machine Learning to Human Rights Data: Two Use Cases from the Bitter Aloe Project Abstract:South Africa’s Truth and Reconciliation Commission (TRC) documented gross human rights violations during apartheid through individual testimonies and incident descriptions. While this approach provided an unprecedented view of political violence, limited planning for future accessibility has constrained researchers to keyword searches and detailed reading of individual transcripts. Bitter Aloe seeks to improve access to the TRC archive by applying two natural language processing techniques: named entity recognition and document embedding. This talk will explore the methodologies used in Bitter Aloe and demonstrate their potential as a broader application of machine learning to human rights…

February 20, 202512:15 PM - 1:15 PM Speaker:Mami Hayashida, ITS-RCI, University of Kentucky and Vikram Gazula, CCS, University of Kentucky Where:327 McVey Hall(Zoom link: https://uky.zoom.us/j/82467171189) Title:Introduction to Retrieval-Augmented Generation Abstract: Despite their impressive performance and versatility, LLMs are fundamentally limited by the data they were initially trained on. For researchers who wish to incorporate LLMs into their work, this often becomes a challenge as the available models trained on general knowledge fail to make reference to the domain knowledge.  Similarly, LLMs do not keep up with the most up-to-date information as each LLM’s knowledge is “frozen”.  RAG (Retrieval Augmented Generation) is a simple technique to overcome this gap by feeding additional knowledge to the LLM as part of workflow. When deployed effectively, the same LLMs will incorporate the added knowledge…