CORA
Cork Open Research Archive (CORA) is UCC’s Open Access institutional repository which enables UCC researchers to make their research outputs freely available and accessible.
UCC Research Communities
Recent Submissions
Setting up a peer-to-peer (P2P) staff listening service at an acute university teaching hospital
(Equinox Publishing Ltd., 2025-07-14) Crane, Sarah; Nuzum, Daniel
This paper is written to describe the establishment and ongoing sustaining of a chaplaincy-led peer-to-peer (P2P) listening service in an acute National Health Service (NHS) Trust in the United Kingdom, which was implemented as a service improvement. It considers the value of compassionate listening within the context of loneliness evidenced in the general population and the documented occupational burden of working in the health service. It reflects on how a peer listening service has been developed and shaped by the role and values of chaplaincy and spiritual care, and delivered by a range of healthcare disciplines and roles. This service has had 6,771 contacts with staff, with 34% being primarily work-related. Recommendations are made for a dedicated team to lead a new service such as this and the positive impact of peer-to-peer listening for healthcare staff wellbeing.
A non-Bayesian preference elicitation approach for noisy decision models
(University College Cork, 2025) Pourkhajouei, Samira; Wilson, Nic; Prestwich, Steve; Science Foundation Ireland; Insight SFI Research Centre for Data Analytics
Interactive preference elicitation plays a crucial role in multi-criteria decisionmaking (MCDM) and recommender systems, particularly when user preferences are uncertain or noisy. Traditional Bayesian approaches offer a principled way to handle such uncertainty by representing preferences as probabilistic models, updating them iteratively based on user responses. However, these methods often entail high computational costs and require prior assumptions on preference distributions, making them less feasible for real-time applications.
This thesis introduces a non-Bayesian framework for interactive preference elicitation under noisy user responses, offering a computationally efficient alternative to probabilistic approaches. Instead of relying on prior distributions, our method identifies and refines a set of plausible preference models by evaluating their consistency with observed user responses. The framework incrementally selects targeted queries that maximize information gain while remaining robust to response inconsistencies, allowing for an effective tradeoff between query efficiency and recommendation accuracy.
We propose a novel query selection strategy that prioritizes alternatives based on their potential optimality across multiple preference models, reducing the number of interactions required to identify the user’s most preferred alternative. To ensure scalability, our method operates on a finite approximation of the preference space, allowing for efficient real-time decision-making.
Empirical evaluations demonstrate the effectiveness of the proposed approach, showing that it achieves high accuracy in identifying user preferences while significantly reducing computational costs compared to Bayesian and minimax regret-based methods. The results suggest that this non-Bayesian framework is well-suited for interactive decision-support systems, personalized recommendations, and real-time MCDM applications, offering a practical balance between efficiency, adaptability, and robustness to noise.
Virtual reality protein visualisation and multiplayer education
(University College Cork, 2024) Xu, Tianshu; Tabirca, Marius-Sabin; Tangney, Mark; Science Foundation Ireland
This research explores the integration of VR in the visualisation and design of protein structures, emphasising its revolutionary impact on educational and research methodologies in bioinformatics and protein design. The research harnesses VR's capability to transform traditional two-dimensional visualisation approaches into interactive, three-dimensional simulations, thereby enhancing the comprehension and manipulation of complex protein models.
A significant part of the thesis is dedicated to developing "Pocket Peptides", a single-player game designed to enhance scientific outreach through interactive protein design. The game uses tools such as I-TASSER for protein structure prediction, UCSF Chimera for three-dimensional model conversion, and Unity and C# for game design and development, integrating these technologies into a gameplay environment that supports educational and research objectives.
Furthermore, the research expands into multiplayer gaming with "Pepblock Builder VR" and "ProMVR", extending VR's educational benefits by enabling collaborative and interactive learning experiences. This multiplayer environment supports the sharing and manipulation of protein designs. It cultivates a learning space where users can engage with and learn from each other in real-time, fostering a sense of community and shared learning.
This study demonstrates VR's potential to enhance the educational landscape of protein engineering significantly, making complex scientific concepts more accessible and engaging through gamified learning environments. The findings suggest that VR can be a powerful tool in the future of scientific education and research, providing immersive experiences that promote a deeper understanding of the protein molecular world, empowering learners to grasp even the most intricate concepts.
Machine learning for structural reasoning in Boolean Satisfiability
(University College Cork, 2024) Dalla, Marco; O'Sullivan, Barry; Visentin, Andrea; Research Ireland Centre for Research Training in Artificial Intelligence; Insight SFI Research Centre for Data Analytics; Science Foundation Ireland
The intersection of machine learning and propositional Boolean Satisfiability (SAT) offers transformative possibilities for solving some of the most challenging computational problems. This thesis investigates the use of modern machine learning (ML) and deep learning (DL) methodologies to enhance Boolean Satisfiability and Model Counting. Building upon the foundational understanding of SAT and its role as an NP-complete problem, this work addresses challenges related to structural reasoning, satisfiability and model counting prediction, feature extraction and instance generation.
The contributions of this research are varied. First, by leveraging a diverse array of features and representations, we develop machine learning algorithms for SAT/UNSAT classification, problem categorization, and approximate model counting. These models demonstrate predictive accuracy and superior computational efficiency compared to traditional handcrafted heuristics. Second, we automate these procedures through the deployment of deep learning algorithms, significantly reducing dependency on manual engineering while improving adaptability to diverse SAT instances. Finally, we extend the application of the deep learning architectures to SAT instance generation. These approaches enable the generation of structurally diverse, statistically realistic SAT instances that serve as robust benchmarks for solver evaluation. The introduction of novel evaluation metrics ensures the practical utility of these generative models, emphasizing their contributions to advancing SAT-solving strategies.
Through an extensive comparative analysis of traditional and machine learning driver SAT-analysis methodologies, this thesis extends the body of work that focuses on the integration of these paradigms. Its findings contribute to the broader field of computational logic, offering insights into scalable and interpretable solutions for Boolean reasoning tasks.
Efficient adaptation of Large Language Models for digital media and government applications
(University College Cork, 2024) Trust, Paul; Minghim, Rosane; Zahran, Ahmed; Science Foundation Ireland
The digital transformation has greatly increased the amount of data, particularly text generated across various fields, including the public sector and digital media. Information such as political campaigns, media reports, citizen feedback, and press releases is now commonly shared on digital platforms. Politicians, government agencies, businesses, and citizens use these platforms to express their views, strategies, goals, and policies.
Analyzing this data can provide valuable insights into public opinion, ongoing policy discussions, business strategies, and socio-political dynamics.
However, the sheer volume of data makes traditional manual analysis or early computational methods impractical, highlighting the need for more efficient automated approaches from Natural Language Processing (NLP), particularly Large Language Models (LLMs), to manage, analyze, and summarize this information. In the course of this work, document-based learning has seen a landmark advance with the progress of generative Artificial Intelligence and the availability of engines and models that are revolutionizing NLP.
Despite significant adoption and investment in the private sector, academia, and high-resource fields, LLMs are less utilized in low-resource fields and the public sector due to constraints such as lack of labeled data, and insufficient budget allocation for machine learning infrastructure and training.
In this Thesis, which was developed during a transition time of fast development of LLMs, we study ways of adapting the most up to date models to novel scenarios in order to both achieve efficiency and understanding of how to adapt LLMs to applications.
To achieve this, this Thesis adapts LLMs to applications in the public sector and digital media. The main approaches developed include: applying weak supervision by leveraging synthetic labels generated by other LLMs to fine-tune models for classifying news articles related to Economic Policy Uncertainty; proposing LLM-based methods for classifying citizen feedback into different categories, as well as for summarization and question answering of citizen feedback; adapting LLMs for automating the handling and navigation of public documents by incorporating strategies such as Retrieval Augmented Generation (RAG), LLM agents; and developing techniques for hallucination detection in these domains.