69238 research outputs found
Sort by
“Ko au mo ʻeku fonua, ʻoku mā taha.” Exploring Tongan student mental well-being and belonging at Waipapa Taumata Rau
No full textThere has been limited research exploring Tongan students’ experiences of mental
well-being and belonging at university. University is a crucial time in students' lives where
they are equipped with skills and knowledge for their future careers. However, university life
presents a unique time of growth and challenges that can result in higher stress levels
compared to non-students. The current qualitative study was guided by the Kakala Research
Framework and aimed to better understand Tongan students’ experiences of mental well-
being and belonging at Waipapa Taumata Rau, University of Auckland.
The present study involved 10 Tongan university undergraduate students (male n = 2;
female n = 8) aged 17–30 years. Data were collected using the Pacific research method of
Talanoa, which facilitated open and reciprocal dialogue between the researcher and
participants. This approach provided a deeper understanding of their experiences. Through
reflexive thematic analysis underpinned by Tongan values, five core themes were identified: 1) the pressure to achieve academic success, 2) intergenerational differences in understanding
well-being, 3) a sense of belonging within Tongan/Pacific communities and spaces, 4) the
challenge of balancing family, church, study, work, and personal life, and 5) recognising self-
care through service. Overall, the findings highlight the unique experiences of Tongan
students, which is crucial for informing policy and support structures at universities that are
responsive to the needs and aspirations of Tongan students
VKORC1 mutations in house mice in the Auckland Region (Aotearoa/New Zealand)
No full textIntroduced house mice are widespread in Aotearoa/New Zealand, and they have significant impacts on native wildlife. The most common toxins for controlling rodents are anticoagulant rodenticides (AR). Even though AR are an efficient tool, resistance to these substances in rodent populations has been detected in many countries. This phenomenon represents a major factor in reducing the success of pest management, and it is mostly related to missense mutations in the VKORC1 gene. Despite the crucial importance of effective house mouse management, genetic AR resistance in mice in Aotearoa/New Zealand is poorly understood. In this study, we undertook a genetic survey of six sites across the Auckland region to investigate the presence of VKORC1 mutations potentially involved in AR resistance. We found a total of five different missense mutations across four of the six sites. Three mutations leading to amino acid changes have been recorded in rodents previously while two are novel. Among these, the well-known Tyr139Cys, involved in resistance to some powerful AR like bromadiolone, is found with a high allelic frequency in central Auckland. Our results suggest that even across a moderate geographic region, there can be important genetic diversity and clustering in AR resistance. Anticoagulant rodenticides are a critical tool in introduced rodent management, but their use must be deliberated and genetic screening of rodent populations should increasingly be an important part of AR management operations
Catalogue of the Greek and Roman Coins in the W.K. Lacey Antiquities Collection
No full textThe W.K. Lacey Antiquities Collection is the teaching collection of ancient
Mediterranean artefacts kept by the Discipline of Classical Studies and Ancient History at Waipapa Taumata Rau – the University of Auckland. The collection is named in honour of Emeritus Professor Walter Kirkpatrick Lacey (1921-2011), who held the chair in Classics at the University of Auckland from 1969 to 1986.
This catalogue, along with the digital record of the collection alongside which it was compiled, seeks to broaden awareness of and access to the Greek and Roman coins in the W.K. Lacey collection to enhance their potential to contribute to numismatically focused research and education
Particle-Trail: Digital Fabrication and Subtractive Manufacturing for Complex Natural Stone Furniture
No full textParticle-Trail is a furniture element generated from a conceptual design created using Voronoi patterns and inspired by the spongy, inner structure of bone. It represents a fusion of the organic with the technology of a real-time 3D-digital development platform. Through algorithmic shaping, Particle-Trail, moves beyond concept to be realized as a travertine architectural element. The project originates from the research question: ‘can the inner structure of bone–which pro-vides critical support and strength to the human skeleton–be transposed and trans-formed into a stone architectural element through subtractive manufacturing?’ Two arch-shaped structures–equal but opposite–become the supporting structure of a table, a modular furniture element with the potential to be scaled for use in different contexts and scenarios, providing a structure of support for distinctive architectural objects. The paper summarizes the design and production process of Particle-Trail as a response to the main research question. A prototype, show-cased at an international exhibition, provided proof of concept for this research. Current stages are investigating the re-use of manufacturing waste generated in the creation of new, circular designs.https://www.sdsbe2024.com
Media emotion intensity and commodity futures pricing
No full textThis study investigates the impact of media emotion intensity on commodities futures returns. Emotion intensity measures the proportion of emotional content relative to factual content in media news. The media emotion intensity factor generates an annual premium of 13% after transaction cost. This premium is more pronounced for commodities with low media coverage, high momentum, high basis-momentum, high hedging pressure, and backwardation. Emotion intensity significantly predicts the trading tendencies of both commercial and non-commercial traders and the cross-section of commodity futures returns at both portfolio and individual levels. We also find that media emotion intensity predicts future commodities’ sentiment. Further, other commonly considered risk sources cannot subsume the predictability of the media emotion intensity factor
It’s Everywhere and Nowhere: The Geological Occurrence and Morphological Characteristics of Erionite in New Zealand
No full textErionite is a fibrous zeolite mineral series linked to malignant mesothelioma, an aggressive and deadly form of cancer. Classified as a Group 1 carcinogen by the IARC, it poses a significant global hazard. However, its characteristics and risks in New Zealand remain poorly understood. This thesis investigates the geological occurrence, morphology, chemistry, and potential toxicity of erionite in New Zealand. Using systematic sampling, advanced mineralogical analyses (e.g., XRPD, SEM-EDS, TEM 3DED, micro-Raman, and EMPA), and in vitro toxicity testing, this study explores erionite’s highly localised distribution and health implications.
From over 100 samples analysed across New Zealand, erionite was identified in nine samples from three regions: Kaipara, Auckland, and Mt. Somers. Morphological and chemical analyses confirmed the presence of erionite-Na, -Ca, and -K species, reflecting their distinct geological origins. Erionite-Na in marine tuffs (Kaipara) forms prismatic aggregates, erionite-K in hydrothermally altered volcanic deposits (Mt. Somers) displays woolly, asbestiform fibres, and erionite-Ca in Auckland occurs as hexagonal bundles of acicular fibres or individual needle-like fibres. Many fibres met WHO respirability criteria, emphasising their potential health hazard if airborne. Cytotoxicity assays revealed that fibre morphology strongly influences toxicity. Erionite from Gawler Downs caused up to 90% cell death over seven days, while Kaipara erionite caused 80% cell death. These results exceeded the toxicity of crocidolite asbestos, with cristobalite in the Kaipara sample likely amplifying its effects.
Identification of erionite in New Zealand presented unique challenges due to its low concentrations and highly localised occurrence. Further complexity arose from the presence of surface coatings, such as clays that artificially inflated Mg and Fe cation contents, complicating accurate chemical compositional analysis. Accurate identification therefore required a suite of advanced techniques. Furthermore, the localised nature of erionite highlights the challenges of systematic identification and emphasises the need for detailed regional investigations. Given erionite’s carcinogenic potential, future research should prioritise its analysis during construction projects, assess long-term exposure pathways, and develop stringent dust management strategies. This research integrates geological, mineralogical, and toxicological perspectives, providing a framework for understanding and managing erionite hazards in New Zealand
Increasing Internet Access for Cochlear Implant Recipients: The Development and Analysis of a Training Programme for Adult Cochlear Implant Recipients
No full textCochlear implants (CIs) have become increasingly effective in addressing substantial auditory deficits. Despite advances in this technology, there are still areas where CI users face significant auditory limitations when compared to individuals with normal hearing. Some of
these limitations, like those associated with communication difficulty, can be addressed with
additional assistive technologies. The digital divide, however, may prevent many adult CI
recipients from fully realizing the benefit of these technologies. It is expected that offering
tailored training to these individuals that addresses their technology related needs will improve their access to communication and their autonomy.
The focus of this study was the development of a training course for adult CI recipients on
the internet and some of its associated technologies. Course structure was based on effective adult learning models that emphasise individualized and accessible training. It aimed to
address barriers that adult CI recipients face in utilizing the internet and technology through
information delivery, collaborative discussions and hands on experience with these tools.
Individual interest’s and skill level were considered in the development of the course content and associated training content package. Questionnaires were utilized before and after
training to assess the outcomes of training and the quality of participant experiences.
Results suggest that the training delivered had an overall positive impact on participants.
Notable improvements were found in measures of self-efficacy, social inclusion and digital access. Participants reported high satisfaction with training indicating that participants felt a positive outcome from training. Possible improvements to course delivery and content were assessed for use in future iterations of this training
Developing And Assessing Language Models For Logical Reasoning Over Natural Language
No full textRecent advancements in AI have highlighted the importance of integrating deep learning with symbolic logic reasoning. Language models such as RoBERTa, DeBERTa, LLaMA, Alpaca, Vicuna, GPT-3.5, and GPT-4 have advanced the performance of AI systems in various natural language processing tasks to human-like levels. However, the generalization of language models in logical reasoning remains underexplored. One of the main reasons is the limitation posed by the lack of extensive, balanced, and real-world datasets for logical reasoning. This thesis has three research objectives, addressing the main research gap/limitation:
1) To improve the models' out-of-distribution performance on multi-step logical reasoning tasks through logic-driven data augmentation.
2) To enhance the models' performance on real-world logical reasoning datasets by constructing an Abstract Meaning Representation based logic-driven data augmentation method.
3) Although large language models demonstrate impressive performance on current logical reasoning leaderboards, it remains underexplored whether they truly possess strong capabilities in logical reasoning.
The first part of the thesis focuses on improving language models' ability in multi-step logical reasoning, particularly when faced with unbalanced reasoning steps. Inspired by DeepLogic, we present IMA-GloVe-GA, an RNN-based model with a gate attention mechanism, developed to accommodate varying reasoning depths. This is facilitated by our PARARULE-Plus dataset, created for deeper reasoning tasks. Our results show notable enhancements in model performance under both standard and out-of-distribution conditions.
The second part of the thesis focuses on generating diverse training data to address the scarcity of real-world logical reasoning datasets and enhance large language models (LLMs) for logical reasoning tasks. We introduce AMR-LDA, a data augmentation method that converts text into Abstract Meaning Representation (AMR) graphs, improving reasoning datasets. This approach benefits various models, including GPT-3.5 and GPT-4, and improves performance, notably achieving the top rank on the ReClor leaderboard.
The third part of the thesis examines how Large Language Models (LLMs) like GPT-3.5 and GPT-4 respond to trivial changes in logical reasoning datasets. We created ReClor-plus, LogiQA-plus, and LogiQAv2-plus, which include shuffled options and modified correct choices to test LLMs' logical reasoning. Although LLMs excel on standard datasets, they exhibit degraded performance with these modified versions. Our findings reveal that incorporating task variations, perturbations in training sets, and logic-driven data augmentation significantly enhances LLMs' generalisation and robustness in logical reasoning scenarios.
This thesis explores several different approaches to demonstrate a more robust QA system that aids computers in thinking and reasoning over natural language texts through logical reasoning. Our methods have been evaluated and now lead the public logical reasoning leaderboard, ReClor. We are the first group in the world to have scored above 90\% on the ReClor hidden test set
Robot Cognitive Architecture with Integrated Memory for Adaptive Human-Robot Collaborative Assembly
No full textThis thesis presents a novel robot cognitive architecture and implementation that enables robots to learn and collaborate with human partners in complex and unstructured environments. While existing human-robot collaboration (HRC) technologies work well in simple setups, they struggle with fluent interaction in complex scenarios when human intentions are difficult to predict and workspace configurations are flexible. This work addresses these limitations through a cognitive architecture built on the SOAR architecture, focusing on internal cognitive processes, including real-time learning, adaptive decision-making, and knowledge evolution. The architecture unifies perception, learning, memory, and execution components, allowing robots to continuously develop skills through human interaction.
The architecture integrates perception capabilities, including object detection and body tracking, with a real-time exploration strategy and a multi-memory system. The perception module provides comprehensive workspace awareness through object detection and human gesture recognition. The exploratory learning mechanism enables autonomous skill acquisition during task execution without requiring offline training phases. The symbolic memory system comprises three interconnected components: semantic, episodic, and procedural memory that collectively support knowledge evolution and enable transparent decision-making through dynamic appraisal mechanisms.
Experimental validation was conducted using the Generic Assembly Box (GAB) to compare the proposed cognitive architecture against three baseline approaches: manual assembly, behavior-based HRC, and learning-based HRC without memory integration. Compared to manual assembly, all HRC methods substantially improved assembly efficiency, with the cognitive system achieving the shortest average completion time. Compared to the behavior-based HRC system, the cognitive architecture achieved a higher task success rate and improved assembly efficiency, reducing task completion time by 13.6% and minimizing human workload with a 23.5% decrease in required interventions. Relative to the learning-based HRC approach, the cognitive system accelerated memory management, as evidenced by a 22.2% reduction in learning iterations needed for stable long-term memory formation. These results collectively highlight the adaptability, efficiency, and interaction fluency of the proposed cognitive architecture, validating its advantages over conventional HRC methods.
These findings demonstrate the system's potential for real-world deployment and suggest promising directions for advancing human-robot collaboration in manufacturing environments
Synthesis of Amphiphilic Polymer Co-Networks via the Growth of a Living RAFT Networks
No full textThis thesis presents a novel approach to synthesizing APCNs via the photo growth of a 'living' parent network. In this study, a hydrophilic parent network was first synthesized through RAFT polymerization, and hydrophobic monomers were inserted into the network through photo growth, forming APCNs with a biphasic structure. The research focused on analyzing the effects of parent network irradiation time on conversion, crosslinking density, and network structure, as well as the influence of photo growth time on the performance of the daughter network. In addition, the study explored the effect of crosslinker ratios on APCN swelling behavior, thermal properties, and mechanical strength. To further validate the advantages of this method, Amphiphilic Copolymer Networks (ACPNs) were synthesized using random copolymerization and compared with APCNs.
Various characterization techniques were used in this study, including Fourier Transform Infrared (FT-IR) spectroscopy to analyze chemical structure, Thermogravimetric Analysis (TGA) to evaluate thermal stability, Scanning Electron Microscopy (SEM) to observe microstructure, and swelling tests to measure APCN swelling performance in polar and non-polar solvents. The results showed that the photo growth method is an efficient and precise synthesis strategy. By optimizing the irradiation time of the parent network and the photo growth time of the daughter network, the ratio of hydrophilic and hydrophobic monomers in APCNs can be effectively adjusted, improving their amphiphilicity, mechanical properties, and structural uniformity. Comparative studies showed that APCNs have more significant phase separation, higher mechanical strength, and better network uniformity compared to ACPNs synthesized by random copolymerization, demonstrating the advantages of the photo growth method in designing amphiphilic networks