31825 research outputs found
Sort by
Mitigating gender bias at the language representation level through latent structure understanding and beyond
No full textAndrews, RickWord embeddings are now a cornerstone in natural language processing (NLP) and machine learning, but they often inherit and perpetuate societal biases, particularly gender bias. This limitation manifests in applied scenarios such as resume screening and machine translation, potentially leading to unfair outcomes. Our research addresses this critical issue by developing advanced techniques for debiasing word embeddings while preserving their semantic utility. ☐ We introduce a novel framework called the Debiasing-assisted Deep Generative Model, which integrates concepts of gender equity into its modeling approach. This model comprises three key building blocks: a feature extraction layer utilizing pre-trained word embeddings and an encoder; a dimension-reduced latent space layer capturing hidden structures and patterns, coupled with a semi-supervised classification layer; and a generative decoder for adaptive re-sampling and reconstruction of debiased embeddings. This architecture employs re-weighting and resampling techniques based on latent structure understanding, coupled with a bias scoring mechanism to mitigate bias effectively while preserving semantic integrity. ☐ Our research begins with the development and evaluation of the De-biasing-assisted Deep Generative Model framework, demonstrating its efficacy in producing debiased word embeddings. The study includes extensive empirical analysis of real-world datasets and user studies to validate the framework’s effectiveness against state-of-the-art benchmarks. We then explore the application of de-biased embeddings in Neural Machine Translation (NMT) systems, addressing the challenge of gender bias when translating from gender-neutral languages. This pioneering work proposes methods for integrating de-biased embeddings into NMT systems, aiming to reduce male default bias in occupations and roles, thus promoting fairer translations. ☐ Finally, we scrutinize the impact of debiased word embeddings on resume job-matching scenarios. This study aims to practically examine the effect of compounding imbalances in real-world resume data and job classifications, evaluate the effectiveness of the Debiasing-assisted Deep Generative Model framework in real-time scenarios for mitigating gender bias in resume screening and job matching, and propose additional techniques to mitigate gender bias by balancing the training data distribution for both genders across different occupations. We analyze how biased representations in job assignments affect classification processes, scrutinizing the nuanced language of resumes and the presence of subtle gender biases. Our findings reveal significant correlations between gender discrepancies in classification true positive rates and gender imbalances across professions. Through these studies, we validate our proposed debiasing techniques, demonstrating their superior performance in mitigating gender bias while maintaining the utility of embeddings for various NLP tasks. This research enriches significantly to the development of gender fairer NLP systems, offering both theoretical advancements and practical implications for promoting gender equity in machine learning-driven applications.University of Delaware, Institute for Financial Services AnalyticsPh.D
Advances in Set Function Learning: A Survey of Techniques and Applications
No full textThis article was originally published in ACM Computing Surveys. The version of record is available at: https://doi.org/10.1145/3715905.
© 2025 Copyright held by the owner/author(s).
This work is licensed under a Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0).Set function learning has emerged as a crucial area in machine learning, addressing the challenge of modeling functions that take sets as inputs. Unlike traditional machine learning that involves fixed-size input vectors where the order of features matters, set function learning demands methods that are invariant to permutations of the input set, presenting a unique and complex problem. This survey provides a comprehensive overview of the current development in set function learning, covering foundational theories, key methodologies, and diverse applications. We categorize and discuss existing approaches, focusing on deep learning approaches, such as DeepSets and Set Transformer-based methods, as well as other notable alternative methods beyond deep learning, offering a complete view of current models. We also introduce various applications and relevant datasets, such as point cloud processing and multi-label classification, highlighting the significant progress achieved by set function learning methods in these domains. Finally, we conclude by summarizing the current state of set function learning approaches and identifying promising future research directions, aiming to guide and inspire further advancements in this promising field.This project is supported in part by National Science Foundation under IIS-2144285 and IIS-2414308
Deepening water scarcity in breadbasket nations
No full textThis article was originally published in Nature Communications. The version of record is available at: https://doi.org/10.1038/s41467-025-56022-6.
© The Author(s) 2025.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.Water is crucial for meeting sustainability targets, but its unsustainable use threatens human wellbeing and the environment. Past assessments of water scarcity (i.e., water demand in exceedance of availability) have often been spatially coarse and temporally limited, reducing their utility for targeting interventions. Here we perform a detailed monthly sub-basin assessment of the evolution of blue (i.e., surface and ground) water scarcity (years 1980-2015) for the world’s three most populous countries – China, India, and the USA. Disaggregating by specific crops and sectors, we find that blue water demand rose by 60% (China), 71% (India), and 27% (USA), dominated by irrigation for a few key crops (alfalfa, maize, rice, wheat). We also find that unsustainable demand during peak months of use has increased by 101% (China), 82% (India), and 49% (USA) and that 32% (China), 61% (India), and 27% (US) of sub-basins experience at least 4 months of scarcity. These findings demonstrate that rising water demands are disproportionately being met by water resources in already stressed regions and provide a basis for targeting potential solutions that better balance the water needs of humanity and nature.L.M. acknowledges the support of the National Science Foundation grant CBET-2144169. K.F.D. and L.M. were supported by the United States Department of Agriculture National Institute of Food and Agriculture grant 2022-67019-37180. M.C.R. and D.D.C. are supported by the NEXUS NESS project founded by the PRIMA Programme, an Art.185 initiative supported and funded under Horizon 2020, the European Union’s Framework Programme for Research and Innovation, with grant agreement no. 2042. D.D.C. was supported by Premio Florisa Melone. Q.D. and W.X. were supported by the National Natural Science Foundation of China (NSFC) (grant nos. 72261147472, 72348003, 72061147001, and 42377467), Ministry of Science and Technology of the People’s Republic of China (grant no. 2023YFD2300301) and National Social Science Foundation of China (grant no. 21BJY013)
Opioid prescribing policies of states with high overdose-related deaths: a 2020 profile
No full textRich, DanielBruch, Sarah K.The current opioid epidemic has been ongoing in the United States since the 1990’s with various policy responses initiated at different levels of government aimed at addressing differing aspects of the epidemic. Over the last two decades, since the early 2000’s, states have taken on the opioid epidemic. This dissertation provides a descriptive portrait of state opioid prescribing policies and analyzes similarities and differences across several key features. The descriptive portrait describes the number and type of policies created; the inclusion of delegated authority, with or without stipulations; the targets of the policies; and the language used to reference opioids. The comparative analysis explores whether and how chronic pain and/or suicide are included in the policy; the policy restrictiveness; and the enforceability of the policy. This project uses a deductive qualitative content analysis of opioid policy documents among states with high opioid death rates. Opioid prescribing policies from the sample states were iteratively coded from which a profile for each state was created and utilized to answer the research questions. A substantial amount of variation in the number and type of policies created was found along with a higher than expected amount of specific stipulations in instances of delegated authority, and variation in the targets of the policies with some states having different prescribing policies dependent on the medical professional. Opioid prescribing policies were more similar than not in terms of inclusion of chronic pain, death by suicide, and enforceability, it was in restrictiveness in which a considerable amount of unexpected dimensionality and variation were discovered.University of Delaware, School of Public Policy and AdministrationPh.D
Sluicing and subject islands: An experimental approach
No full textThis article was originally published in Glossa: A Journal of General Linguistics. The version of record is available at: https://doi.org/10.16995/glossa.11054.
Glossa: a journal of general linguistics is a peer-reviewed open access journal published by the Open Library of Humanities. © 2025 The Author(s). This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License (CC-BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. See http://creativecommons.org/licenses/by/4.0/.This paper investigates how sluicing in English interacts with islandhood (Ross, 1967), focusing on how the subject island constraint applies under both regular sluicing (with an indefinite correlate) and contrast sluicing (with a focused correlate). We conducted three experiments. Experiment 1 examined what the structure of the elided clause is, by having participants choose a possible continuation. We found different patterns for regular sluicing versus contrast sluicing, with contrast sluicing exhibiting a strong preference for syntactic parallelism with the antecedent clause. Regular sluicing, in contrast, preferred a non-parallel continuation (a copular clause), consistent with evasion approaches to island repair under sluicing (e.g., Barros et al., 2014). Next, Experiment 2 examined the sensitivity of sluicing to island effects. Strikingly, contrast sluicing did not demonstrate notable acceptability degradations when subject islands were involved (contra predictions made by Barros et al., 2014; Griffiths & Lipták, 2014; Merchant, 2008), but we did find an overall degradation in contrast sluicing in comparison to regular sluicing. Because this outcome challenges previous assertions regarding the sensitivity of contrast sluicing to island constraints, Experiment 3 asked whether the subject island effect is truly at play in embedded wh questions. We found that – like with matrix questions but unlike with relative clauses (Abeillé et al., 2020) – subject island effects do emerge in embedded interrogatives. Since evasion is not a
possible approach for contrast sluicing, whatever the source of the subject island effect is, it disappears when the relevant structure is not pronounced
Development of transition-metal catalyzed methods utilizing silicon-containing compounds and nitroalkanes
No full textWatson, Donald A.Organosilanes and nitroalkanes are important synthetic intermediates in organic synthesis. Nitroalkanes can be transformed to medicinally relevant functional groups, such as tertiary amines. Organosilanes can participate in C–C bond formation via the Hiyama-Denmark cross-coupling or be converted to alcohols via the TamaoFleming oxidation. Silicon bioisosteres are also of interest in the medicinal chemistry community. Despite their significance, general methods to synthesize organosilanes, particularly those that are chiral, and nitroalkanes have proven challenging. Our group uses transition metal-catalyzed strategies to access these interesting compounds. ☐ In Chapter 1, I will describe a Hiyama-Denmark cross-coupling of tetrasubstituted vinylsilanes to stereospecifically access tetrasubstituted alkenes, which are an important motif in medicinal chemistry and organic synthesis. A general method was developed to cross couple a variety of highly substituted vinylsilanes with aryl halides under mild conditions using KOSiMe3 as the base and THF/DMA as the solvent, without the need for any extraneous additives. The identification of dimethyl- (5-methylfuryl)vinylsilanes as an easily synthesized, bench stable, yet reactive coupling partner was essential. Mechanistic investigations and byproduct analysis have revealed a unique and unexpected role for DMA, namely that it serves as a reagent for the slow release of water into the solution over the course of the reaction. Details of the optimization, scope, and the investigations that led to our mechanistic understanding are described. ☐ In Chapter 2, the development of a light-mediated nickel-catalyzed Calkylation of nitroalkanes will be described. The direct photoactivation of nickel complexes to enable catalysis has only recently been explored and is not well documented for C–C bond formation reactions. In previous work from our group, a dual photoredox nickel-catalyzed C-alkylation of nitroalkanes was reported. Control experiments revealed the nickel complex can catalyze the reaction without the need for a rare-earth metal photocatalyst. The scope of this transformation has been extensively studied and a benchtop protocol for this transformation has been developed.University of Delaware, Department of Chemistry and BiochemistryPh.D
Moth community composition in urbanized landscapes
No full textTallamy, Douglas W.Few studies have compared the community composition of urban and exurban moth faunas, and fewer still have sought to quantify their difference in biomass. Moths are critical to the functioning of terrestrial ecosystems, including as a main food resource for most terrestrial bird species. I trapped adult moths at 15 sites along an urban to forest gradient in the Boston, Massachusetts, US area to determine the effects of urbanization on moth community composition, diversity, and biomass. I found that small forest fragments well under 1 km2 in area contained densities of small- to medium-sized moths similar to those in larger forests, whereas large woody plant-feeding moths occurred mainly in larger forests. My findings indicate that urban woodlots dominated by native plants are comparable to forests in their density of moth biomass except for large species. This is a hopeful sign that urban habitat restoration has greater potential conservation value than is often assumed.University of Delaware, Department of Entomology and Wildlife EcologyM.S
Illusion Rainsdrops- a 4D Knitting Exploration
No full textThis article was originally published in International Textile and Apparel Association Annual Conference Proceedings. The version of record is available at: https://doi.org/10.31274/itaa.18821.
© 2024 The author(s). Published under a Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.When combining 3D knitted textures and shapes with optical illusion patterns, an interactive 4D effect can be achieved, an interdisciplinary research area of growing interest for the author and many other scholars. One optical illusion knitting technique lesser known originated with a Japanese teacher who explained a simple technique of alternating rows of dark and light-colored yarns knitted with a garter stitch to produce patterns that appear and disappear depending on the angle from which the fabric is viewed. The purpose of this project was to explore how the illusion knitting technique can be manipulated into a shaped garment using various 3D knitting techniques for an optical illusion slimming 4D effect. This scholarship, combining techniques from old craftmanship with current knitting technologies, elevates the use of illusion knitting from purely aesthetic means to a functional 4D effect aimed at optically sliming the silhouette
Non-Isothermal Melt Crystallization of a Biodegradable Polymer Studied by Two-Dimensional Infrared Correlation Spectroscopy
No full textThis article was originally published in Molecules. The version of record is available at: https://doi.org/10.3390/molecules30051131.
© 2025 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).The non-isothermal melt crystallization process of poly[(R)-3-hydroxybutyrate-co-(R)-3-hydroxyhexanoateate] (PHBHx) was monitored using attenuated total reflection infrared (ATR IR) measurement. The resulting time- and temperature-dependent spectra were subjected to the two-dimensional correlation spectroscopy (2D-COS) analysis. The C=O stretching region of the PHBHx sample consisted of several distinct IR contributions attributable to the population of amorphous component, well-ordered type I lamellar crystal, and less ordered inter-lamellar type II crystal. The spectral intensity change in type I crystal occurs in the earlier stage of the crystallization at a higher temperature range compared to the overall intensity decrease in the amorphous component occurring throughout the crystallization process. The growth of the type II crystal started in a later stage at a lower temperature than the creation of the type I crystal. An early decrease in a small but distinct portion of the amorphous component may be related to a crystallization precursor species with some level of molecular order. Hetero-mode correlation analyses revealed that the crystalline band intensity changes in the C-H stretching and fingerprint regions all occur later than the population changes in crystalline species reflected by the carbonyl stretching bands. This observation suggests that the spectral intensity changes in the C-H stretching and fingerprint regions do not directly represent the population dynamics of the crystalline and amorphous species but probe instead the molecular state of the crystalline entities still undergoing the evolutionary changes.This research received no external funding
Development and characterization of disulfide-based self-healing sealants for concrete pavement joints
No full textTatar, JovanMany highways and bridges built decades ago are now reaching the end of their service life, making advancements in material performance increasingly important. Self-healing materials have the potential to autonomously repair damage, thereby enhancing material durability. The development of self-healing polymers so far has been primarily focused on strategies to achieve self-healing, with less attention given to the application-related aspects of these materials. However, without specific material performance requirements, it is challenging to identify future steps for improving the design of self-healing polymers. This study explores the potential of using self-healing sealants for joints in concrete pavements. The sealant incorporates a thiol-epoxy system, with tertiary amines as initiators of the polymerization reaction. The base of the sealant is a liquid polysulfide (Thiokol). The polysulfide component, rich in disulfide bonds, is theoretically capable of undergoing exchange reactions to enable self-healing. The effects of cross-linking, thiol-epoxy stoichiometry, and initiator type were investigated to establish correlations between polymer structure, polymer properties, and self-healing capabilities. This analysis also served as a parametric study where the aim was to determine which formulation best satisfies the specific application requirements. The study addresses uncertainties related to the functionality of commercially available thiols and the interference of epoxy homopolymerization with the thiol-epoxy addition mechanism, particularly in the presence of nucleophilic tertiary amines. Self-healing behavior was characterized by performing tensile testing experiments to quantify the recovery of strength and elongation after self-healing under ambient conditions. In addition, stress-relaxation experiments at different temperatures were carried out to understand the mechanisms driving self-healing. To determine the potential of using triethylamine as a catalyst for disulfide exchange at room temperature, a model reaction between low molecular weight disulfides was examined. The performance aspects of self-healing polymers were evaluated, emphasizing key metrics such as mechanical properties, adhesion to concrete, the impact of mechanical forces on healing efficiency, and resistance to environmental degradation. Findings from this study contribute to the implementation of autonomous self-healing polymers with dynamic disulfide bonds in construction plastics to increase their service life and reduce environmental impact.University of Delaware, Civil, Construction and Environmental EngineeringPh.D