Knowledge graphs (KGs) are a powerful way of representing information for digital humanities. However, non-technical users often struggle at the outset of exploration, a challenge defined as the Initial Exploration Problem. The Tús Maith framework addresses this issue through curated natural language questions and answers (CuQAs) created from Competency Questions (CQs) that aim to convey the scope of a KG and provide meaningful entry points into it. While prior work has explored using large language models (LLMs) for CQ template generation, the template-filling step, where questions and answers are instantiated with entity information, remains a key challenge. In this paper, we evaluate whether LLMs have the capacity to support domain experts in this stage, focusing on the Virtual Record Treasury of Ireland (VRTI) KG, where accuracy, provenance, and robustness are crucial for practical use. Using structured JSON inputs derived from popular search terms and expert-authored templates, we generated and assessed 24,900 question-answer pairs across four LLMs (GPT-5, DeepSeek-V3.1, Gemini 2.0 Flash, Qwen-2.5-72B) under two provenance conditions (basic vs. full). Our evaluation considers slot fidelity, semantic similarity, completeness, hallucination rates, and runtime efficiency, with statistical tests conducted per run per LLM, and additional batch-level analysis (n = 68) to isolate provenance requirement effects. We further show that a lightweight JSON validation check is an effective proxy for ground truth semantic evaluation of factual question-answer pairs. These LLM-generated, validated questions form an intermediate step in the lifecycle from abstract CQ templates to filled-in questions and answers intended to be reviewed and refined by the VRTI KG’s domain experts (historians) to produce the final user-facing questions (CuQAs). To demonstrate the practical impact, we present a prototype (TMv1) of the Tús Maith framework and highlight the design implications for curator-facing interfaces: provenance-transparent interaction, validation-integrated workflows, and performance-transparent model selection.

Provenance refers to the details about the entities, activities, and individuals involved in the creation of a piece of data or an object, which can help assess its quality, reliability or trustworthiness. This paper describes an ontology that can be used to represent provenance and contextual information about Patient Generated Health Data (PGHD) shared with an Electronic Health Record (EHR). We created the ontology leveraging existing standard ontologies. Further, we chose and modeled important concepts based on literature that captures contextual information relevant to clinical decision making. We illustrated our ontology using a use case from literature and evaluated the ontology using competence questions. In the future, it will be useful to employ this ontology in translating the PGHD collected for clinical decision making.

Existing sentiment analysis models have yet to succeed in recognizing the clinical sentiment polarity in intensive care unit clinical notes. Conversely, natural language processing techniques have been through significant advancement and multiple paradigm shifts recently, enhancing the feasibility of improving clinical sentiment tasks on one hand, but increasing the difficulty in selecting the most effective approach. To address this shortcoming, this research empirically investigates existing sentiment analysis models when applied to critical care clinical notes. It seeks to provide insights on which approach has the potential to improve the quality of clinical sentiment recognition within this domain-specific frame. In this study, we compare nine selected models from three families: lexicon models, BERT models, and prompt models, and across the general and clinical domain. We compare the task on whole notes truncated at default limit by language models against the section of notes including assessment section onwards. Our dataset uses a proxy for sentiment, based on timing of notes relative to positive or negative outcomes such as discharge or death. The analysis concludes that the best performing models are Scifive, and clinicalBERT with accuracy ranges (0.80 - 0.93) depending on the type of notes, and that the worst performing models were Textblob and AFINN lexicons with accuracy ranges (0.48 - 0.54) depending on the type of notes. Based on our experiments, our proxy for sentiment is sufficiently accurate to support this line of experimentation, and segments of notes with assessments and plans provide comparable results, with slight improvement, to default truncation of whole notes.

Patient-Generated Health Data (PGHD) holds promise for enhancing personalized healthcare, but its diverse origins pose trust challenges for clinicians. This article outlines the collaborative creation of the PGHD Trust Canvas, a tool devised to help clinicians evaluate the trustworthiness of PGHD in their decision-making processes. Developed through an iterative design process and informed by interviews with nine clinicians from various disciplines, the PGHD Trust Canvas draws inspiration from similar existing models like the Business Model Canvas, the Trust Canvas and the Ethics Canvas. It offers a structured approach for clinicians to assess PGHD systematically, aiming to facilitate its appropriate integration into clinical decisions. This initiative represents a significant step toward leveraging PGHD safely and effectively in healthcare.

This article explores the concept of patient-generated health data (PGHD) and its potential to be used in clinical decision-making. PGHD is data created and recorded by patients outside of clinical settings, such as through wearable technology or mobile apps. While PGHD has the potential to provide valuable insights into patients' health and behaviour, little is known about the perspective of clinicians when facing with the decision whether to trust or distrust PGHD. 13 clinicians of various disciplines, including nurses and doctors were recruited using the snowball sampling method. Qualitative methods were used in this study to analyse the results. The results suggest that some aspects of Quality, Provenance and Risk are important to clinicians and that there is a need for guidance and governance to assist clinicians in this decision.

Infectious diseases are a major cause of human morbidity, but most are avoidable. An accurate and personalized risk prediction is expected to alert people to the risk of getting exposed to infectious diseases. However, as data and knowledge in the epidemiology and infectious diseases field becomes available, an updateable risk prediction model is needed. The objectives of this article are (1) to describe the mechanisms for generating a Bayesian Network (BN), as risk prediction model, from a knowledge-base, and (2) to examine the accuracy of the prediction result. The research in this paper started by encoding declarative knowledge from the Atlas of Human Infectious Diseases into an Infectious Disease Risk Ontology. Automatic generation of a BN from this knowledge uses two tools (1) a Rule Converter generates a BN structure from the ontology (2) a Joint & Marginal Probability Supplier tool populates the BN with probabilities. These tools allow the BN to be recreated automatically whenever knowledge and data changes. In a runtime phase, a third tool, the Context Collector, captures facts given by the client and consequent environmental context. This paper introduces these tools and evaluates the effectiveness of the resulting BN for a single infectious disease, Anthrax. We have compared conditional probabilities predicted by our BN against incidence estimated from real patient visit records. Experiments explored the role of different context data in prediction accuracy. The results suggest that building a BN from an ontology is feasible. The experiments also show that more context results in better risk prediction.

Argumentation theory (AT) has been gaining momentum in the health care arena thanks to its intuitive and modular way of aggregating clinical evidence and taking rational decisions. The basic principles of argumentation theory are described and demonstrated in the breast cancer recurrence problem. It is shown how to represent available clinical evidence in arguments, how to define defeat relations among them and how to create a formal argumentation framework. Argumentation semantics are then applied over the built-framework to compute arguments justification status. It is demonstrated how this process can enhance the clinician decision-making process. A encouraging predictive capacity is compared against the accuracy rate of well-established machine learning techniques confirming the potential of argumentation theory in health care.

Clinical guidelines are central to the implementation of clinical decision support systems (CDSSs). Addition or revision of clinical guidelines usually causes the (re-) development of new or existing CDSSs. The separate maintenance of clinical knowledge and their driving systems implies extra system development cost and low knowledge delivery efficiency. We propose, in this paper, an approach to liaise the two activities and support a complete knowledge-driven CDSS architecture. It will accommodate and disseminate new knowledge with minimum efforts required to make relevant changes to the systems, but make use of the new knowledge whenever it becomes available. A Multi-Agent System architecture and a rule-based knowledge repository are put together to realize this goal.

The World Wide Web (WWW) provides access to a vast array of educational content, a great deal of which is ideal for incorporation into eLearning experiences. However sourcing, harvesting and incorporating appropriate content has proven to be a complex and arduous task. This paper introduces a system that enables the discovery and classification of educational content from open corpus sources, such as the WWW, and facilitates the incorporation of such content into eLearning systems. The open corpus content service (OCCS) discovers, harvests and indexes content through the implementation of a focused Web crawler, content classifier and indexer. This reduces the cognitive load placed on the educator by content authoring, allowing them to focus on the pedagogical design of eLearning offerings.