Dartmouth Institute for Health Policy and Clinical Practice
Dartmouth Digital Commons (Dartmouth College)Not a member yet
8214 research outputs found
Sort by
Revitalization Of Endangered Languages With AI
Full text linkThe preservation and revitalization of endangered languages, particularly those with minimal digital presence, presents significant challenges for computational linguistics. This thesis addresses these challenges by proposing novel methods for language identification and data generation, focusing on underrepresented Indigenous languages, specifically Nüshu, Native American and Native Alaskan languages.
In the first study, a COLING 2025 paper, we present NüshuRescue, an AI-driven framework designed to facilitate the preservation of Nüshu, an endangered script used exclusively by Yao women in China. Using minimal seed data, we demonstrate how GPT-4-Turbo can generate new translations, expanding a publicly available Nüshu-Chinese corpus, achieving 48.69% accuracy in translating unseen examples.
The second study, a NAACL 2025 paper, introduces a Random Forest classifier tailored for identifying Native American languages, including Navajo, which has been consistently misidentified by existing language identification tools like Google’s LangID. By leveraging a custom dataset and training on misidentified languages, we achieve near-perfect classification accuracy, illustrating the potential of lightweight, decentralized language identification models.
Finally, in the third study, an ACL 2025 Findings paper, we extend this work to 20 Native Alaskan languages, using few-shot prompting with LLMs and fine-tuning with XLM-RoBERTa to achieve near-perfect identification accuracy for these endangered languages. These results underscore the feasibility of building robust language technologies for low-resource languages with minimal data, contributing to both the technical field and the broader efforts to preserve and revitalize Indigenous languages. This thesis provides a scalable approach to endangered language identification, with the potential to make a significant impact on linguistic diversity preservation in the digital age
A HIGH SPEED IMAGING FRAMEWORK USING SPARSE SAMPLING AND GENERATIVE PRIORS
No full textHigh-speed imaging systems are used to study the motion of particles in fluid dynamics, capture shockwave propagation, and analyze the behavior of materials under stress. Observing such rapid phenomena allows for a deeper understanding of complex systems that otherwise are invisible to the human eye. However, traditional imaging systems face significant challenges when capturing high-speed events. These systems are often constrained by bandwidth limitations, preventing them from acquiring data at the necessary rates to capture fast-moving objects with the required spatial detail. We incorporate generative AI to reconstruct the images to overcome bandwidth limitations. By conditioning these models on the limited captured data, we reconstruct the images, with a primary focus on spatial resolution. We use a single-pixel detector capable of acquiring data at rates up to 2 gigahertz (2 billion measurements per second) for high-speed data capture. This is combined with a high-speed galvanometer mirror system, enabling fast scanning, specifically following Lissajous curves. The system design focuses on maximizing data capture within the bandwidth constraints while leveraging sparse measurements to reconstruct spatial images. Although current work has demonstrated image reconstruction from sparse data for static images, the approach holds promise for reconstructing dynamic events in the future. This research paves the way for next-generation high-speed imaging systems capable of achieving spatial and temporal resolution
Futurismo Cósmico: Sonic Dislocation Under the Fifth Sun
No full textFuturismo Cósmico is a text in which I tune into the interstices of my sonic practice: popular musical styles, alternative tuning systems, and different, contrasting epochs of my experience as Mexicano and Latino. I discuss three projects of mine that engage with sound, time, and space in different ways. The journey concludes by looking at my practice in the context of a culturally dislocated space I inhabit, in my own psyche. However, understanding that I am a social being, this same practice is what provides a drive and a sense of purpose in an otherwise loosely articulated identity
All Things Pass Away
Full text link“All Things Pass Away” is a work of crime fiction. Set amidst the downfall of local print journalism, laid off sportswriter Mark McGinnis travels from Rochester, New York to the oil boomtown of Fort McMurray, Alberta to pursue a freelancing project on troubled NHL draft prospect Noah Jeannot. Once there, McGinnis’s profile on the player is thrown into question when Jeannot disappears after an apparent violent altercation near his billet family’s home. Exploring themes of nostalgia, ambition, and obsession, the novella traces McGinnis’s attempts to understand not only the culture of the twin gods – oil and hockey – which rule the town but also the various suspects within it, each with a vested interest in Jeannot’s career. Finding himself enmeshed in the investigation, McGinnis’s efforts to produce a cohesive story are thwarted in a landscape riddled with self-interest – that is, until he learns to play their game. In doing so, he discovers the myriad ways in which one’s past remains a part of them, only capable of being overcome through the crucible of experience, introspection, and re-invention
The Magdala Moms of La Jolla
No full textThe Magdala Moms of La Jolla is a satirical novel set at The Magdala School, an elite Christian academy in La Jolla, California, where social status, curated faith, and community performance are indistinguishable. The story follows Mitzi Madigan, a brash, ambitious mother whose carefully constructed persona hides a troubling past. Her world begins to unravel when Caroline—a single mother working in construction—is granted a scholarship for her daughter, Emma, to attend Magdala.
As Emma enters the private school’s polished social ecosystem, she begins to uncover the secrets buried beneath the sun-drenched surface. The Mag Moms, a clique of hyper-polished, scripture-quoting women who pride themselves on appearances and piety, quickly close ranks. Emma’s presence threatens the illusion of unity and control they\u27ve worked hard to maintain.
Told through alternating perspectives, flashbacks, and a slow-burning detective thread, the novel reveals how religious language is used to mask judgment, exclude outsiders, and perpetuate privilege. The tone balances biting humor with emotional insight, exposing the rot beneath righteousness while treating its characters with layered humanity.
The result is a work that explores themes of belonging, identity, and performative faith in a community where even grace can be weaponized. The Magdala Moms of La Jolla uses fiction to satirize the power structures embedded in elite religious subcultures, asking who is allowed in, who is kept out, and what must be sacrificed to stay
You\u27re Such A Princess
Full text linkYou’re Such a Princess is a five-play anthology that reimagines well-known folk tales. By re-imagining four Brothers Grimm tales and one by Hans Christian Andersen, I follow a centuries-old tradition. As scholar Donald Haase explains, folk tales evolve with each retelling, allowing their ‘morals’ to shift over time. I have created encounters between real people (extending characters beyond Vladimir Propp’s Dramatis Personae) and revised the endings to challenge the audience. My Snow White character hews closer to the pejorative implications of calling someone a ‘princess’ in our society: entitled, arrogant, spoilt. Two of my storylines (Little Red Cap and The Little Mermaid) revert closer to their original versions emphasizing female agency and desires, which differ significantly from their respective Brothers Grimm and Disney versions. All of my renditions highlight the care of friends and family in creating a happy ending. My interventions may subvert dominant paradigms (e.g., heteronormative marriage) but also deal more broadly with contemporary issues. The spindle prick of Sleeping Beauty translates to the opioid crisis, and the transcendence of sea foam in The Little Mermaid translates into the fading away of an eating disorder. While the thesis addresses the one-dimensionality of the ‘princess’ archetype (all the protagonists are young women), the role of ‘prince’ is also subverted in ways designed to explore masculinity in our society. Rather than simply being a trophy to be won, men have vulnerabilities, dreams and power issues, ranging from self-medicating for mental health problems to a reliance on transactional sexual power. I chose playwriting for this storytelling, both as a return to the oral tradition but also to underscore the tradition of allowing future interpretations. Stagecraft is malleable and can render visual key themes, for instance a gigantic mirror divides the stage in Snow White to emphasize that the heroine mirrors the villain. Song, both original and copyrighted, features throughout to convey emotional tone. In the Little Mermaid, it goes further in representing song as the main character’s ‘voice’/agency and joy. I have written these plays to reflect contemporary mores that I believe should be reinforced, particularly the importance of human connection
Computational Modeling and Structural Generation of Piano Music in the Classical Style
Full text linkListening to fast-tempo piano sonatas of the Classical period (circa 1750-1820) has been shown to have therapeutic effects for neurological disorders such as epilepsy. The limited existing repertoire of music in this style motivates the creation of more long-form, coherent compositions with clearly defined structure. Despite the long history of computer-based music generation and recent progress in deep learning, particularly transformer-based models, generating structurally coherent long-form music remains a major challenge. This difficulty stems from the scarcity of reliable structural annotation datasets, the computational demands of modeling very long musical sequences, and the lack of effective structural encoding in both symbolic music representations and model architectures.
The goal of this thesis is to design, implement, and evaluate a framework for generating coherent Classical sonata-style piano music, with a particular focus on extended duration and well-defined structural organization. The main contributions of this work are as follows: (i) AutoStruc, an unsupervised structural annotation system that integrates multi-modal music representations to construct phrase-section structural annotations from Classical piano sonatas; (ii) REMI-Lite, a novel symbolic music tokenization scheme that emphasizes a hierarchical organization of beat, bar, phrase; (iii) a generative framework that models the compositional process of Classical piano sonatas by assembling phrase-level components---generated by two transformer-based architectures, nextGEN and accGEN---into coherent, well-structured sections or movements; and (iv) task-specific evaluation methods, including a novel objective metric, Melody Retrieval, to assess model correctness, and behavioral studies to evaluate the musicality, rhythmic regularity, and structural coherence of the generated compositions.
To our knowledge, this is the first attempt to apply transformer-based architectures with a strong emphasis on structural modeling for the generation of Classical piano sonatas
Computation-Driven Design and Synthesis of Molecular Sensors for Surface-Enhanced Raman Spectroscopy in Liquids
No full textThe reliable detection of molecules in liquid environments is a continuing challenge in chemical sensing and biological diagnostics. While surface-enhanced Raman spectroscopy (SERS) is well-recognized for its high sensitivity and molecular specificity, conventional platforms – based on colloidal nanoparticles or patterned substrates – have shortcomings. These conventional configurations struggle with signal reproducibility, non-uniform hotspot distribution, and, critically, restricted accessibility for large targets such as proteins and exosomes.
This thesis develops a computation-guided approach to developing next-generation SERS sensors capable of overcoming these persistent obstacles. At the heart of the study is the design and fabrication of metal-insulator-metal (MIM) nanoparticles, where a metallic core and surface metal nanoparticles are separated by a dielectric spacer. Finite element analysis (FEA) is used to simulate and optimize the electromagnetic field enhancements achieved by different MIM structures, with a focus on maximizing both signal intensity and molecular accessibility. These simulations provide clear guidelines for the synthesis of nanoparticles with controlled core size, dielectric spacer thickness, and surface nanoparticle arrangement.
The synthesis process involves both experiments and theoretical models, including the application of Derjaguin-Landau-Verwey-Overbeek (DLVO) theory to guide controlled MIM nanoparticle assembly and minimize nanoparticle aggregation. The resulting MIM nanoparticles demonstrate robust SERS performance for both small molecules and large biomolecular analytes in liquid samples. Protocols for surface modification and liquid-phase detection are developed, ensuring that large targets can interact effectively with the sensor’s enhanced fields.
In parallel, this work establishes robust data analysis strategies that combine spectral preprocessing with statistic digitalization and machine learning techniques. These methods help ensure consistent quantification and accurate identification of complex and heterogeneous SERS signals.
In summary, the research presented here offers a solution to current limitations in SERS-based sensing. By integrating computation, synthesis, and data analytics, this thesis aims to advance SERS technology in molecular diagnostics and other real-world applications
Strategies to improve antibody-mediated protection against herpes simplex virus infections
Full text linkNo efficacious vaccine exists for herpes simplex virus and the global burden of disease remains high. Antibodies can serve as potent therapeutics to protect high risk populations such as neonates. Although rare, neonatal HSV (nHSV) infections can result in significant mortality and lifelong neurological morbidity even with antiviral therapy. While HSV-specific maternally derived antibodies significantly reduce the risk of neonatal disease in humans, the mechanisms by which they mediate protection have yet to be described. Using a mouse model of neonatal HSV (nHSV) infections we aimed to determine the mechanisms of antibody-mediated protection using monoclonal antibodies (mAbs) targeting two of the major surface glycoproteins involved in viral entry, glycoproteins D and B (gD, gB).
Direct administration of HSV-specific mAbs to neonatal mice protected them from HSV-1 and HSV-2-mediated mortality. mAbs against gD required both viral neutralization and Fc effector functions for broad and potent protection against HSV-1. For protection against HSV-2, however, Fc function alone, and not neutralization was required for these gD-specific mAbs. In contrast to gD-specific mAbs, mAbs targeting gB required effector functions for protection against HSV-1 in neonatal mice. Viral neutralization, while contributing to protection, was not the dominant mechanism of protection in this model. mAb Fc- and IgG subclass engineering also demonstrated that rendering IgG Fc insensitive to viral immune evasion by HSV gE/gI improved mAb efficacy in vitro and in vivo. This engineering strategy improved mAbs targeting both gD and gB, demonstrating the broad applicability to improve mAb therapeutics against HSV infections. In sum, these studies suggest the particular importance of effector functions in mediating protection against HSV infection in neonatal mice and that Ab-engineering represents a promising path forward to improve mAb-mediated protection
Aestheticizing and Engendering Disease: The Case of Tuberculosis
No full textThis thesis examines how tuberculosis, known in the nineteenth century as “consumption,” was aestheticized in American cultural and medical discourse, particularly in industrial cities of the North between 1830 and 1850. It addresses the paradox at the heart of the disease’s representation: while tuberculosis was medically pathologized, it was simultaneously romanticized, especially when afflicting elite white women. The central problem explored is how this aestheticization of illness functioned to reinforce gender, class, and racial hierarchies by transforming suffering into a symbol of virtue, refinement, and social status. Using a cultural-historical methodology, this project analyzes a range of sources, including medical treatises, women’s magazines, advertisements, fashion critiques, and advice literature. Through close analysis and contextual interpretation, it demonstrates how medical rhetoric and consumer culture worked in tandem to produce and circulate the image of the consumptive woman as a desirable and morally elevated figure. It also draws on the work of scholars such as Susan Sontag, Clark Lawlor, Michel Foucault, and Carolyn Day to situate the American consumptive aesthetic within broader transatlantic and metaphorical frameworks. The findings reveal that tuberculosis became a flexible cultural metaphor aligning fragility, moral sensitivity, and elite femininity with the visible symptoms of disease. While white, upper-class women were permitted to suffer beautifully in medical and visual culture, working-class and racialized individuals were denied access to this ideal. Their suffering was framed as pathological, unsightly, and morally suspect. Moreover, the thesis shows that physicians were not detached observers but active participants in consumer culture, selling remedies, endorsing fashion norms, and contributing to a commodified vision of health. Tuberculosis thus emerges not only as a medical condition, but as a cultural script through which social power was negotiated. The consumptive woman exemplifies how nineteenth-century American society transformed illness into a tool for reinforcing ideals of femininity, class distinction, and white superiority. By foregrounding the role of consumer culture in shaping medical and aesthetic discourses, this thesis repositions tuberculosis at the center of antebellum debates about identity, morality, and social order. The consumptive ideal offered women symbolic value only through suffering and decline, a paradox that affirmed subordination even in moments of cultural visibility