Conference Papers
Permanent URI for this collectionhttps://mro.massey.ac.nz/handle/10179/7616
Browse
86 results
Search Results
Item Mining Meaningful Keys and Foreign Keys with High Precision and Recall(VLDB Endowment, 2025-01-01) Koehler H; Link S; Palpanas T; Tatbul NWe demonstrate a next-generation Entity/Relationship (E/R) Profiler that mines meaningful key/foreign key relationships from a given data repository. Core novelties include a strict hierarchy of key variants ranging from candidate keys to SQL unique constraints that represent different ways to identify incomplete entities, a measure of orthogonality that separates accidental from meaningful keys, and algorithms for mining approximate keys for all these variants under different thresholds of arity, completeness, dirtiness, and orthogonality. We showcase the high precision and recall achieved by our tool and how it facilitates the users’ understanding which entity and referential integrity constraints govern their data.Item Mātauranga Moana: uplifting Māori and Pacific values of conceptualisation over western co-design constructs(Design Research Society, 2023-11-29) Withers S; Stokes G; Jones D; Borekci N; Clemente V; Corazzo J; Lotz N; Nielsen LM; Noel L-AThis paper offers a critical examination of the problematic use of western co-design methodologies when applied to indigenous and diasporic communities. By centring place-based, relational design approaches to enable cultural conventions from our position in Aotearoa New Zealand, we argue the use of co-design constructs risks overlaying neo-liberal ideologies on top of our resilient indigenous Māori and Pacific knowledge systems, values, ethics, and collective approaches towards design conceptualisation. As design researchers located in te moana-nui-a-Kiwa our discussion is underpinned by our Māori whakapapa, Sāmoan gafa, and relationship to Te Tiriti o Waitangi. We present our kōrero through a case study relationship with a local healthcare service, aiming to increase access for Māori and Pacific tamariki through design actions. Our collaboration was developed within the format of a tertiary course involving Māori and Pacific tauira enrolled in Design and Fine Arts degrees at Ngā Pae Māhutonga School of Design, Te Kunenga ki Pūrehuroa Massey University of New Zealand. Unlike traditional university design courses that aim to achieve a specific measurable outcome, we focussed on fostering whakawhānaungatanga and evidencing this through activated learning of the cultural conventions of wānanga and talanoa towards weaving together our values through critically reflective practice. Our case study relationship demonstrates the importance of relational place-based knowledge systems and their conditions for enabling reflexivity towards tino rangatiratanga and ola manuia within Māori and Pacific communities; further highlighting the systemic barriers that practices of co-design can seed when attempting to serve our communities in Aotearoa.Item A Framework to Assess Multilingual Vulnerabilities of LLMs(Association for Computing Machinery, 2025-05-23) Tang L; Bogahawatta N; Ginige Y; Xu J; Sun S; Ranathunga S; Seneviratne SLarge Language Models (LLMs) are acquiring a wider range of capabilities, including understanding and responding in multiple languages. While they undergo safety training to prevent them from answering illegal questions, imbalances in training data and human evaluation resources can make these models more susceptible to attacks in low-resource languages (LRL). This paper proposes a framework to automatically assess the multilingual vulnerabilities of commonly used LLMs. Using our framework, we evaluated six LLMs across eight languages representing varying levels of resource availability. We validated the assessments generated by our automated framework through human evaluation in two languages, demonstrating that the framework's results align with human judgments in most cases. Our findings reveal vulnerabilities in LRL; however, these may pose minimal risk as they often stem from the model's poor performance, resulting in incoherent responses.Item Learning to Bound for Maximum Common Subgraph Algorithms(Schloss Dagstuhl – Leibniz-Zentrum für Informatik, Dagstuhl Publishing, Germany, 2025-08-08) Kothalawala BW; Koehler H; Wang Q; Garcia de la Banda MIdentifying the maximum common subgraph between two graphs is a computationally challenging NP-hard problem. While the McSplit algorithm represents a state-of-the-art approach within a branch-and-bound (BnB) framework, several extensions have been proposed to enhance its vertex pair selection strategy, often utilizing reinforcement learning techniques. Nonetheless, the quality of the upper bound remains a critical factor in accelerating the search process by effectively pruning unpromising branches. This research introduces a novel, more restrictive upper bound derived from a detailed analysis of the McSplit algorithm's generated partitions. To enhance the effectiveness of this bound, we propose a reinforcement learning approach that strategically directs computational effort towards the most promising regions within the search space.Item Multi-lingual mathematical word problem generation using long short term memory networks with enhanced input features(European Language Resources Association (ELRA), 2020-01-01) Liyanage V; Ranathunga SA Mathematical Word Problem (MWP) differs from a general textual representation due to the fact that it is comprised of numerical quantities and units, in addition to text. Therefore, MWP generation should be carefully handled. When it comes to multi-lingual MWP generation, language specific morphological and syntactic features become additional constraints. Standard template-based MWP generation techniques are incapable of identifying these language specific constraints, particularly in morphologically rich yet low resource languages such as Sinhala and Tamil. This paper presents the use of a Long Short Term Memory (LSTM) network that is capable of generating elementary level MWPs, while satisfying the aforementioned constraints. Our approach feeds a combination of character embeddings, word embeddings, and Part of Speech (POS) tag embeddings to the LSTM, in which attention is provided for numerical values and units. We trained our model for three languages, English, Sinhala and Tamil using separate MWP datasets. Irrespective of the language and the type of the MWP, our model could generate accurate single sentenced and multi sentenced problems. Accuracy reported in terms of average BLEU score for English, Sinhala and Tamil languages were 22.97%, 24.49% and 20.74%, respectively.Item Concentration of 12 Oligosaccharides in the Milk of New Zealand Breastfeeding Women(MDPI (Basel, Switzerland), 2023-03-23) Jia LL; Brough L; Weber JL; Smith C; Mackay S; Jalili-Moghaddam S; Gibbs MHuman milk oligosaccharides (HMOs) are the third most abundant component in breast milk. HMOs benefit infant gut health, modulate immune responses, and promote brain development. The profile and concentration of HMOs vary considerably among breastfeeding women, and are reported to be associated with genetic, maternal, and environmental factors as well as feeding practices. One reason for the diversity in HMO concentration is the secretor gene, which determines the presence of an enzyme responsible for the synthesis of 2′-FL and LNFP-I. To date, there is no report about HMO concentration or profile in the New Zealand population. Our objective was to investigate 12 HMO concentrations in a small sample of New Zealand women. Sixty-eight breastfeeding mothers (mean age 32 years, 77% Caucasian) of singleton infants (median age [Q1, Q3] 108 [70, 166] days) were included, with 65% exclusively breastfeeding and 54% who had two or more children. Concentrations of 12 HMOs were measured by UHPLC with fluorescence detection. Overall, 68% of mothers were secretors, which was defined by the presence of 2′-FL in the milk. HMO profiles varied widely; total HMO concentration varied 4.2-fold between women; and individual HMOs varied from 4.8-fold to >100-fold. The median of total HMO concentration (Q1, Q3) of the secretors and non-secretors were 6774.9 (6395.4, 8245.6) mg/L and 7128.0 (6093.1, 7880.1) mg/L respectively. Significant differences in concentration of 2′-FL, 3-FL, A-Tet, LNFP-I, LNFP-II, LNFPV, and LNnT between secretors and non-secretors were found by Mann–Whitney tests. However, there was no significant difference in concentrations of LNFP-III, LNnFP, 3′-SL, 6′-SL, LNT, or total HMOs between the secretors and the non-secretors. HMO concentrations vary broadly between breastfeeding women. A longitudinal cohort of a larger sample size is required to fully investigate HMO profiles at different lactation stages of New Zealand women and to further explore the influence of maternal and environmental factors on HMO concentration.Item Efficacy of needle and endoscopic lavage on the recuperation of microspheres from the adult equine metacarpo−/metatarsophalangeal joint and digital flexor tendon sheath(Wiley Periodicals LLC on behalf of American College of Veterinary Surgeons, 2025-06-25) Beggan CP; Panizzi L; Oliver LJObjectives: To measure microsphere recovery following needle-through-and-through lavage (NTAT) of the metacarpo−/metatarsophalangeal joint (fetlock) and digital flexor tendon sheath (DFTS) compared to endoscopic lavage (EL). Study design: Ex vivo experimental study. Animals: Adult equine cadavers immediately following euthanasia (n = 10). Methods: Colored 15 μm microspheres (2 million) were injected into fetlock joints and DFTS. Synovial structures were assigned to NTAT or EL groups. Each lavage was performed using 5 L of 0.9% NaCl, sequentially collecting egress fluid for microsphere quantification. Recovery was compared using a full-factorial general linear model. Results: There was a significant effect of the liter of egress fluid and microsphere recovery in both fetlocks (p <.01) and DFTS (p <.01), with most microspheres recovered in the first 2 L (79%–83%) for both techniques. More microspheres were recovered in the first liter using NTAT than EL (p <.01) in both fetlocks (659 883 ± 20 820 vs. 567 601 ± 24 452) and DFTS (644 341 ± 17 460 vs. 550 637 ± 38 022). No difference in total recovered microspheres was observed between NTAT lavage of fetlock (981 600 ± 46 839) and DFTS (957 419 ± 45 729) across 5 L (p =.88). Conclusion: Needle-through-and-through lavage was more effective than EL at recovering microspheres in the first liter from cadaveric equine fetlock joints and DFTS. Both techniques demonstrated comparable efficacy between fetlock and DFTS in microsphere recovery following increased lavage volumes. Clinical significance: Needle-through-and-through lavage (NTAT) is a viable alternative for suspected synovial contamination when EL is delayed or not feasible. This study does not evaluate NTAT's efficacy for treating established sepsis or removing pannus/foreign bodies.Item Word embedding evaluation for Sinhala(European Language Resources Association, 2020-01-01) Lakmal D; Ranathunga S; Peramuna S; Herath I; Calzolari N; Béchet F; Blache P; Choukri K; Cieri C; Declerck T; Goggi S; Isahara H; Maegaard B; Mariani J; Mazo H; Moreno A; Odijk J; Piperidis SThis paper presents the first ever comprehensive evaluation of different types of word embeddings for Sinhala language. Three standard word embedding models, namely, Word2Vec (both Skipgram and CBOW), FastText, and Glove are evaluated under two types of evaluation methods: intrinsic evaluation and extrinsic evaluation. Word analogy and word relatedness evaluations were performed in terms of intrinsic evaluation, while sentiment analysis and part-of-speech (POS) tagging were conducted as the extrinsic evaluation tasks. Benchmark datasets used for intrinsic evaluations were carefully crafted considering specific linguistic features of Sinhala. In general, FastText word embeddings with 300 dimensions reported the finest accuracies across all the evaluation tasks, while Glove reported the lowest results.Item Application of absolute sustainability assessment to new zealand residential dwellings(IOP Publishing Ltd, 2020-11-20) McLaren SJ; Chandrakumar C; Dowdell D; Bullen L; Jaques ROne approach to supporting the implementation of sustainable activities by industry sectors is the use of climate targets. Such climate targets have potential to be used in design and rating tools for buildings and to support government regulation for the building and construction sector. In this study, the climate targets for New Zealand residential dwellings were calculated based on assigning the global carbon budget (for limiting temperature increase to 1.5 or 2.0 °C during 2018-2050) to three building typologies: detached, medium-density housing and apartments. These budgets were assigned to the pre-existing and new-built dwellings using building stock projections for the nominated period. Separately, the climate impact of new-built dwellings in each of the three residential typologies were assessed using Life Cycle Assessment methodology. For New Zealand residential buildings, new-built dwellings exceed their 1.5 °C climate targets by a factor of 6.7, 6.8 and 10.9 for detached, medium-density housing, and apartments respectively. For the 2.0 °C climate target, these factors are 4.8, 4.8 and 7.7 for detached, medium-density housing, and apartments respectively. The results show that about two-thirds of the climate impact of residential dwellings for the period 2018-2050 is associated with preexisting dwellings rather than new-builds. The operational energy used for space heating, water heating, lighting and plug loads makes the biggest contribution to the climate impact for all typologies of pre-built residential dwellings. For new-built residential dwellings, both the operational energy and the construction materials/products contribute most of the climate impact.Item Stable Tree Labelling for Accelerating Distance Queries on Dynamic Road Networks(OpenProceedings.org, 2024-11-11) Koehler H; Farhan M; Wang QFinding the shortest-path distance between two arbitrary vertices is an important problem in road networks. Due to real-time traffic conditions, road networks undergo dynamic changes all the time. Current state-of-the-art methods incrementally maintain a distance labelling based on a hierarchy among vertices to support efficient distance computation. However, their labelling sizes are often large and cannot be efficiently maintained. To combat these issues, we present a simple yet efficient labelling method, namely Stable Tree Labelling (STL), for answering distance queries on dynamic road networks. We observe that the properties of an underlying hierarchy play an important role in improving and balancing query and update performance. Thus, we introduce the notion of stable tree hierarchy which lays the ground for developing efficient maintenance algorithms on dynamic road networks. Based on stable tree hierarchy, STL can be efficiently constructed as a 2-hop labelling. A crucial ingredient of STL is to only store distances within subgraphs in labels, rather than distances in the entire graph, which restricts the labels affected by dynamic changes. We further develop two efficient maintenance algorithms upon STL: Label Search algorithm and Pareto Search algorithm. Label Search algorithm identifies affected ancestors in a stable tree hierarchy and performs efficient searches to update labels from those ancestors. Pareto Search algorithm explores the interaction between search spaces of different ancestors, and combines searches from multiple ancestors into only two searches for each update, eliminating duplicate graph traversals. The experiments show that our algorithms significantly outperform state-of-the-art dynamic methods in maintaining the labelling and query processing, while requiring an order of magnitude less space.
