RUSLAN MITKOV
Lancaster University
Large Language Models are still lagging behind their Deep Learning counterparts. Will Prompt Engineering change this reality?
Deep Learning and, more recently, Large Language Models (LLMs) have revolutionised the field of artificial intelligence and have been successfully employed in many disciplines, capturing widespread attention and enthusiasm.
The invited speaker has been closely monitoring the performance of both Deep Learning models and LLMs. Through a series of studies he has led, Mitkov has found that domain-specific Deep Learning models often outperform general-purpose LLMs, a conclusion supported by other researchers in the field. However, a recent study focusing on Holocaust data, conducted by Mitkov’s PhD student Isuri Nanomi Arachchige with Mitkov’s involvement and collaboration with other colleagues, has revealed a new development: effective Prompt Engineering can enable LLMs to surpass high-performing Deep Learning models.
In this presentation, Mitkov will review the results of several studies comparing the performance of Deep Learning models and LLMs across various Natural Language Processing (NLP) tasks. He will then delve into the findings of the recent study, highlighting how prompt engineering can significantly enhance LLMs, potentially giving them an edge over Deep Learning models.
BIOGRAPHY
Ruslan Mitkov is Professor of Computing and Communications at Lancaster University, specialising in Natural Language Processing (NLP), Computational Linguistics and Translation Technology. Before joining Lancaster University, he worked at the University of Wolverhampton, where he set up and led the Computational Linguistics Research Group and was Director of the Institute for Research in Information and Language Processing as well as the Responsible Digital Humanities Lab.
Dr Mitkov has published over 280 peer-reviewed papers and is known for his important contributions to anaphora resolution, computer-assisted generation of multiple-choice tests and the development of next-generation translation memory systems. He is a pioneer in the use of NLP tools to support people with autism.
He is sole editor of the Oxford Handbook of Computational Linguistics and author of the book Anaphora Resolution. He is Executive Editor of the Natural Language Engineering journal of Cambridge University Press and Editor-in-Chief of the John Benjamins NLP book series.
Professor Mitkov has supervised over 30 PhD theses and over 40 Master’s dissertations. He has been a keynote speaker at over 230 international conferences. Conferences and events in the 2023-2024 academic year include the Translation Forum in Riyadh, the ICON’2023 conference in Goa, the KaniTamil2024 conference in Chennai, the UNESCO-organized JIAMCATT in Paris, and several others in Spain, South Africa, Poland, and Bulgaria.
Ruslan Mitkov is a Fellow of the Alexander von Humboldt Foundation in Germany, a Marie Curie Fellow, a Distinguished Visiting Professor at the University of Franche-Comté in Besançon, France, and at the University of Malaga, Spain. He designed and leads the first and only Erasmus Mundus Master’s Program in Technology for Translation and Interpreting (EM TTI), an innovative program with a strong focus on research and business, involving leading companies in the global translation and language industry.
In September 2022, the US National Board of Medical Examiners awarded him a certificate of distinguished collaboration, recognizing his impact on the organization’s strategic planning and decision-making using NLP solutions over the past 17 years.
For his outstanding professional and research achievements, Professor Mitkov has been honored with the title of Doctor Honoris Causa three times: at the University of Plovdiv in 2011, at the University of Veliko Tarnovo in 2014, and at the New Bulgarian University in 2022.
With the sponsorship of:
SHOMIR WILSON
Pennsylvania State University
Natural Language Processing for Privacy Empowerment and Fairness
Natural language processing (NLP) is a core part of our information society, with an enormous variety of impacts. However, privacy risks and harmful biases are concerns when NLP is embedded in sociotechnical systems. I will share my lab’s research to use NLP to push in a positive direction, toward privacy empowerment and fairness for technology users. Recent results of this work include PrivaSeer (https://privaseer.ist.psu.edu/), a search engine and corpus that represent over 1M website privacy policies available and explorable for privacy stakeholders. I will also describe our work to identify sociodemographic biases in popular language models, showing the need for careful attention to the diversity of users’ experiences when developing human language technologies.
BIOGRAPHY
Shomir Wilson is an Assistant Professor in the College of Information Sciences and Technology at the Pennsylvania State University, where he leads the Human Language Technologies Lab. His research interests span natural language processing, privacy, and computational social science. He is particularly interested in breaking down technology’s “walls of text”, situations where a human reader is expected to consume a large quantity of text to take action while lacking time or expertise to properly understand it.
He holds over $2M in active grants from the National Science Foundation and the National Institutes of Health, covering research on usable privacy, legal text, and social justice in law enforcement.
Prior to becoming faculty he held postdoctoral positions in Carnegie Mellon University’s School of Computer Science and the University of Edinburgh’s School of Informatics. He received his Ph.D. in Computer Science from the University of Maryland in 2011.
To learn more about his work, visit https://shomir.net/.
With the sponsorship of Plan de Transformación de la Universidad de La Rioja.
