The hyperfine interaction between carrier spins and nuclear spins is an important component in exploring spin-dependent properties in materials with strong spin orbit interaction.However hyperfine interaction has been less studied in bismuth (Bi), a heavy element exhibiting a strong Rashba-like spin-orbit interaction in its two-dimensional surface states due to the broken spatial inversion symmetry. In this dissertation we experimentally explore the carrier spin polarization due to transport under strong spin-orbit interaction and the nuclear polarization resulting from the relatively unexplored hyperfine interaction on Bi(111) films.The carrier and nuclear spin polarizations are expected to dynamically interact, a topic with ramifications to other materials where surface states with noteworthy properties play a role.To achieve this goal, an optimized van der Waals epitaxy growth technique for Bi(111) on mica substrates was developed and used, resulting in flat Bi surfaces with large grain sizes and a layered step height of 0.39±0.015 nm, corresponding to one Bi(111) bilayer height. A comparison between Bi(111) films grown on three different substrates (mica, InSb(111)B, and Si(111)) is discussed, for which scanning electron microscopy and atomic force microscopy are applied to obtain the structural and morphological characteristics on the film surface. Magnetotransport measurements are carried out to extract the transport properties of theBi(111) films. Using the high quality Bi(111) film deposited on mica, we develop quantum magnetotransport techniques as delicate tools to study hyperfine interaction. The approach is based on measuring quantum corrections to the conductivity due to weak antilocalization, which depend on the coherence of the spin state of the carriers. The carrier spin polarization is generated by a strong DC current in the Bi(111) surface states (here called the Edelstein effect), which then induces dynamic nuclear polarization by hyperfine interaction. Quantum transport antilocalization measurements in the Bi(111) thin-films grown on mica indicate a suppression of antilocalization by the in-plane Overhauser field from the nuclear polarization, and allow for the quantification of the Overhauser field, which is shown to depend on both polarization duration and the DC current magnitude. Various delay times between the polarization and the measurement result in an exponential decay of the Overhauser field, driven by relaxation time T1. We observe that in the Bi surface states, the appreciable electron density and strong spin-orbit interaction allow for dynamic nuclear polarization in the absence of an external magnetic field. / Doctor of Philosophy / This dissertation focuses on the heavy element bismuth (Bi), a semimetal with strong spin-orbit interaction at its two-dimensional surface. Given the challenge to grow high qualityBi(111) films, we present an optimized van der Waals epitaxy technique to grow Bi(111)films on mica substrates, which show a flat surface with large grain sizes and a layered step height of 0.391±0.015 nm, corresponding to one Bi(111) bilayer height. To demonstrate the high quality of the Bi(111) surface, a comparison of surface morphology was conducted among Bi(111) films deposited on three different substrates (mica, Si(111), and InSb(111)B),along with a comparison between their electronic transport properties. By applying a DC current on the high quality Bi(111) film on mica, a carrier spin polarization is established via mainly what we here call the Edelstein effect, which then induces dynamic nuclear polarization by hyperfine interaction and generates a non-equilibrium nuclear spin polarization without externally applied magnetic field. We quantified the Overhauser field from the nuclear polarization all-electrically by conducting quantum transport antilocalization experiments, which showed a suppression of antilocalization by the in-plane Overhauser field.Comparative measurements indicated that the magnitude of the Overhauser field depends onthe spin-polarizing DC current magnitude and the polarization duration. The experiments also show that antilocalization forms a sensitive probe for hyperfine interaction and nuclear polarization.

The advent of social media has revolutionized human communication, significantly improving individuals' lives. It makes people closer to each other, provides access to enormous real-time information, and eases marketing and business. Despite its uncountable benefits, however, we must consider some of its negative implications such as online harassment and cyberbullying. Cyberbullying is becoming a serious, large-scale problem damaging people's online lives. This phenomenon is creating a need for automated, data-driven techniques for analyzing and detecting such behaviors. In this research, we aim to address the computational challenges associated with harassment-based cyberbullying detection in social media by developing machine-learning framework that only requires weak supervision. We propose a general framework that trains an ensemble of two learners in which each learner looks at the problem from a different perspective. One learner identifies bullying incidents by examining the language content in the message; another learner considers the social structure to discover bullying. Each learner is using different body of information, and the individual learner co-train one another to come to an agreement about the bullying concept. The models estimate whether each social interaction is bullying by optimizing an objective function that maximizes the consistency between these detectors. We first developed a model we referred to as participant-vocabulary consistency, which is an ensemble of two linear language-based and user-based models. The model is trained by providing a set of seed key-phrases that are indicative of bullying language. The results were promising, demonstrating its effectiveness and usefulness in recovering known bullying words, recognizing new bullying words, and discovering users involved in cyberbullying. We have extended this co-trained ensemble approach with two complementary goals: (1) using nonlinear embeddings as model families, (2) building a fair language-based detector. For the first goal, we incorporated the efficacy of distributed representations of words and nodes such as deep, nonlinear models. We represent words and users as low-dimensional vectors of real numbers as the input to language-based and user-based classifiers, respectively. The models are trained by optimizing an objective function that balances a co-training loss with a weak-supervision loss. Our experiments on Twitter, Ask.fm, and Instagram data show that deep ensembles outperform non-deep methods for weakly supervised harassment detection. For the second goal, we geared this research toward a very important topic in any online automated harassment detection: fairness against particular targeted groups including race, gender, religion, and sexual orientations. Our goal is to decrease the sensitivity of models to language describing particular social groups. We encourage the learning algorithm to avoid discrimination in the predictions by adding an unfairness penalty term to the objective function. We quantitatively and qualitatively evaluate the effectiveness of our proposed general framework on synthetic data and data from Twitter using post-hoc, crowdsourced annotation. In summary, this dissertation introduces a weakly supervised machine learning framework for harassment-based cyberbullying detection using both messages and user roles in social media. / Doctor of Philosophy / Social media has become an inevitable part of individuals social and business lives. Its benefits, however, come with various negative consequences such as online harassment, cyberbullying, hate speech, and online trolling especially among the younger population. According to the American Academy of Child and Adolescent Psychiatry,1 victims of bullying can suffer interference to social and emotional development and even be drawn to extreme behavior such as attempted suicide. Any widespread bullying enabled by technology represents a serious social health threat. In this research, we develop automated, data-driven methods for harassment-based cyberbullying detection. The availability of tools such as these can enable technologies that reduce the harm and toxicity created by these detrimental behaviors. Our general framework is based on consistency of two detectors that co-train one another. One learner identifies bullying incidents by examining the language content in the message; another learner considers social structure to discover bullying. When designing the general framework, we address three tasks: First, we use machine learning with weak supervision, which significantly alleviates the need for human experts to perform tedious data annotation. Second, we incorporate the efficacy of distributed representations of words and nodes such as deep, nonlinear models in the framework to improve the predictive power of models. Finally, we decrease the sensitivity of the framework to language describing particular social groups including race, gender, religion, and sexual orientation. This research represents important steps toward improving technological capability for automatic cyberbullying detection.

This study investigates the impact of an online community's control structure on the knowledge sharing process in that community. Using a framework comprised of legitimate peripheral participation theory and the weak-ties phenomenon, the study focuses on a comparative analysis of self-organizing online communities (e.g., weblog networks) and centralized online communities (e.g., discussion forums communities) with respect to the efficacy of knowledge sharing in these communities. The findings of this study indicate that self-organizing communities of practice have more weak-ties among their members compared to centralized communities. As per weak-ties theory of Granovetter (1973, 1983), these findings suggest that self-organizing communities facilitate greater dissemination of knowledge and flow of information among their members than centralized communities. The abundance of weak-ties in their community structure also makes self-organizing communities better environments for the discovery of new information compared to centralized community environments. This study did not find any evidence of community structure impact on peripheral participation and the interaction activity level among peripheral participants of a given online community. These observations may have stemmed from the limitations of research design, however, it is safe to say as of now that verdict on peripheral participation differences in different community structures is inconclusive at best. / Master of Science

Three full-scale experimental tests were conducted to investigate the strength and stiffness of weak-axis moment end-plate connections. Each test consisted of two girders connected to a column web with four-bolt extended moment end-plates. Two tests were conducted with bare steel. One test included a composite concrete slab that confined the top extension of the end-plate. Finite element models of the tests were created with the commercial software SAP2000. A simplified modeling procedure was developed to overcome the contact problems between the end-plates and column web, and between the bolts and holes in the end-plates and web. The simplified modeling procedure accurately predicted the experimental elastic stiffness, in the form of column web rotations, of the connections. Yield line theory was used to investigate the plastic strength of the column web. Several yield line patterns were examined. Analytical plastic moment strengths were very conservative when compared to the observed behavior of the column web. The experimental stiffness of the test with the concrete slab confining the top extension of the end-plate was compared to the stiffness of a similar test without a slab. The slab increased the elastic stiffness of the connection; however, after the concrete began cracking and crushing around the connection, the stiffness was greatly decreased. / Master of Science

The study of quantum interference in solid-state systems yields insight in fundamental properties of mesoscopic systems. Electron quantum interference constitutes an important method to explore mesoscopic physics and quantum decoherence. This dissertation focuses on two-dimensional (2D) electron systems in $delta-$Si doped n-type In$_{0.64}$Ga$_{0.36}$As/In$_{0.45}$Al$_{0.55}$As, 2D hole systems in Si-doped p-type GaAs/Al$_{0.35}$Ga$_{0.65}$As and C-doped p-type GaAs/\Al$_{0.24}$Ga$_{0.76}$As heterostructures. The low temperature experiments study the magnetotransport of nano- and micro-scale lithographically defined devices fabricated on the heterostructures. These devices include a single ring interferometer and a ring interferometer array in 2D electron system, Hall bar geometries and narrow wires in 2D hole systems. The single ring interferometer yields pronounced Aharonov-Bohm (AB) oscillations with magnetic flux periodicity of h/e over a wide range of magnetic field. The periodicity was confirmed by Fourier transformation of the oscillations. The AB oscillation amplitude shows a quasi-periodic modulation over applied magnetic field due to local magnetic flux threading through the interferometer arms. Further study of current and temperature dependence of the amplitude of the oscillations indicates that the Thouless energy forms the measure of excitation energies giving quantum decoherence. An in-plane magnetic field was applied to the single ring interferometer to study the Berry's phase and the Aharonov-Casher effect. The ring interferometer array yields both AB oscillations and Altshuler-Aronov-Spivak (AAS) oscillations, the latter with magnetic flux periodicity of h/2e. The AAS oscillations require time-reversal symmetry and hence can be used to qualify time-reversal symmetry breaking. More importantly, the fundamental mesoscopic dephasing length associated with time-reversal symmetry breaking under applied magnetic field, an effective magnetic length, can be obtained by the analysis of the AAS oscillations over magnetic field. A theoretical model for confined ballistic system is confirmed by experimental data fitting. The AAS oscillations are barely resolved above 0.16 T and their amplitude decays with increasing magnetic field. The AB oscillations exist till above 2 T and their amplitude doesn't show the monotonic decay with increasing magnetic field. The different behavior of the AAS and AB oscillations originates in the different symmetries, respectively temporal and spatial, that they are sensitive to. The p-type 2D GaAs system has strong spin-orbit interaction (SOI). Antilocalization in a Hall bar geometry was analyzed by the 2D Hikami-Larkin-Nagaoka (HLN) theory to obtain the spin coherence time and phase coherence time. The 2D hole systems we studied have low density and high mobility, quite different from the 2D electron systems. These high-quality 2D hole systems demonstrate semi-classical ballistic phenomena in mesoscopic structures preferentially to quantum-coherence phenomena. / Ph. D.

Energy harvesting from the rotating system has been an influential topic for researchers over the past several years. Yet, most of these harvesters are linear resonance-based harvesters whose output power drops dramatically under random excitations. This poses a serious problem because a lot of vibrations in rotating systems are stochastic. In this dissertation, a novel energy harvesting strategy for rotating systems was proposed by taking advantage of stochastic resonance. Stochastic resonance is referred to as a physical phenomenon that is manifest in nonlinear bistable systems whereby a weak periodic signal can be significantly amplified with the aid of inherent noise or vice versa. Stochastic resonance can thus be used to amplify the noisy and weak vibration motion. Through mathematical modeling, this dissertation shows that stochastic resonance is particularly favorable to energy harvesting in rotating systems. The conditions for stochastic resonance are satisfied by adding a nonlinear bistable energy harvester to the rotating system because whirl noise and periodic signalㄴ already coexist in the rotating environment. Both numerical and experimental results show that stochastic resonance energy harvester has higher power and wider bandwidth than linear harvesters under a rotating environment. The dissertation also investigates how stochastic resonance changes for the various types of excitation that occur in real-world applications. Under the non-gaussian noise, the stochastic resonance frequency is shifted larger value. Furthermore, the co-existence of the vibrational and stochastic resonance is observed depending on the periodic signal to noise ratio. The dissertation finally proposed two real applications of stochastic resonance energy harvesting. First, stochastic resonance energy harvester for oil drilling applications is presented. In the oil drilling environment, the periodic force in rotating shafts is biased, which can lower the efficacy of stochastic resonance. To solve the problem, an external magnet was placed above the bi-stable energy harvester to compensate for the biased periodic signal. Energy harvester for smart tires is also proposed. The passively tuned system is implemented in a rotating tire via centrifugal force. An inward-oriented rotating beam is used to induce bistability via the centrifugal acceleration of the tire. The results show that larger power output and wider bandwidth can be obtained by applying the proposed harvesting strategy to the rotating system. / Doctor of Philosophy / In this dissertation, a novel energy harvesting strategy for rotating systems was proposed by taking advantage of stochastic resonance. Stochastic resonance is referred to as a physical phenomenon that is manifest in nonlinear bistable systems whereby a weak periodic signal can be significantly amplified with the aid of inherent noise or vice versa. Stochastic resonance can thus be used to amplify the noisy and weak vibration motion. Through mathematical modeling, this dissertation shows that stochastic resonance is particularly favorable to energy harvesting in rotating systems.Both numerical and experimental results show that stochastic resonance energy harvester has higher power and wider bandwidth than linear harvesters under a rotating environment. The dissertation also investigates how stochastic resonance changes for the various types of excitation that occur in real-world applications. The dissertation finally proposed two real applications of stochastic resonance energy harvesting. First, stochastic resonance energy harvester for oil drilling applications is presented. Energy harvester for smart tires is also proposed. The results show that larger power output and wider bandwidth can be obtained by applying the proposed harvesting strategy to the rotating system.

Recent adaptations of machine learning models in many businesses has underscored the need for quality training data. Typically, training supervised machine learning systems involves using large amounts of human-annotated data. Labeling data is expensive and can be a limiting factor in using machine learning models. To enable continued integration of machine learning systems in businesses and also easy access by users, researchers have proposed several alternatives to supervised learning. Weak supervision is one such alternative. Weak supervision or weakly supervised learning involves using noisy labels (weak signals of the data) from multiple sources to train machine learning systems. A weak supervision model aggregates multiple noisy label sources called weak signals in order to produce probabilistic labels for the data. The main allure of weak supervision is that it provides a cheap yet effective substitute for supervised learning without need for labeled data. The key challenge in training weakly supervised machine learning models is that the weak supervision leaves ambiguity about the possible true labelings of the data. In this dissertation, we aim to address the challenge associated with training weakly supervised learning models by developing new weak supervision methods. Our work focuses on learning with constraint-based weak supervision algorithms. Firstly, we will propose an adversarial labeling approach for weak supervision. In this method, the adversary chooses the labels for the data and a model learns by minimising the error made by the adversarial model. Secondly, we will propose a simple constrained based approach that minimises a quadratic objective function in order to solve for the labels of the data. Next we explain the notion of data consistency for weak supervision and propose a data consistent method for weakly supervised learning. This approach combines weak supervision labels with features of the training data to make the learned labels consistent with the data. Lastly, we use this data consistent approach to propose a general approach for improving the performance of weak supervision models. In this method, we combine weak supervision with active learning in order to generate a model that outperforms each individual approach using only a handful of labeled data. For each algorithm we propose, we report extensive empirical validation of it by testing it on standard text and image classification datasets. We compare each approach against baseline and state-of-the-art methods and show that in most cases we match or outperform the methods we compare against. We report significant gains of our method on both binary and multi-class classification tasks. / Doctor of Philosophy / Machine learning models learn to make predictions from data. In supervised learning, a machine learning model is fed data and corresponding labels for the data so that the model can learn to predict labels for new unseen test data. Curation of large fully supervised datasets is expensive and time consuming since it involves subject matter experts providing labels for each individual data example. The cost of collecting labels has become one of the major roadblocks for training machine learning models. An alternative to supervised training of machine learning models is weak supervision. Weak supervision or weakly supervised learning trains with cheap, and easy to define signals that noisily label the data. We refer to these signals as weak signals. A weak supervision model combines various weak signals to produce training labels for the data. The key challenge in weak supervision is how to combine the different weak signals while navigating misleading correlations in their errors. In this dissertation, we propose several algorithms for weakly supervised learning. We classify our methods as constraint-based weak supervision since weak supervision is provided as constraints to our algorithms. We use experiments on different text and image classification datasets to show that our methods are effective and outperform competing methods that we compare against. Lastly, we propose a general framework for improving the performance of weak supervision models by incorporating a few labeled data. With this method we are able to close the gap to supervised learning without the need for labeling all the data examples.

I dagens digitaliserade värld använder 4,45 miljarder sociala medier. Utav dessa befinner sig 120 miljoner registrerade användare på plattformen OnlyFans. Under coronapandemin växte plattformen explosionsartat vilket ledde till en spridning av information om OnlyFans i populärkultur och i media. Vårt syfte med denna kandidatuppsats är att studera informationsspridningen av OnlyFans via sociala medier. Utifrån denna kunskap vill vidare belysa hur unga kvinnors uppfattning om plattformen påverkas av denna sorts spridning av information. Baserat på teorierna weak ties och eWOM (elektronisk word-of-mouth) har vi tagit fram tre indikatorer för att skapa förståelse om unga kvinnors attityd gentemot OnlyFans. Dessa är attityd, övertygelse och trovärdighet.  Information om OnlyFans sprids främst via sociala medier, inte minst av influencers. Med en feministisk ansats har vi med en metodkombination utav kvalitativa intervjuer och en kvantitativ enkät tagit del av unga kvinnors tankar och åsikter om plattformen. Resultatet visar en negativ attityd av den övervägande majoriteten av respondenterna. Med en redan negativ attityd visar respondenterna en svag övertygelse och trovärdighet mot plattformen och dess kreatörer. En intervju med en kreatör från plattformen har visat oss en glimt av OnlyFans som arbetsgivare. Slutligen, baserat på resultaten och analysen, har slutsatsen dragits att respondenterna har en negativ attityd, trovärdighet och övertygelse gentemot plattformen. Detta påverkar inte deras syn på kvinnan som kreatör utan mot den betalande mannen bakom skärmen.

This thesis introduces Comquest, an adaptive framework designed for the large-scale collection and integration of user comments from the Web. User comments are featured on many websites and there is growing interest in mining and studying user comments in applications, such as opinion mining and information diffusion. However, crawling user comments generally requires hard-coded solutions that are tethered to specific websites, which is hard to scale and maintain. To achieve a generalizable and scalable comment crawling solution, Comquest employs two website-agnostic approaches for comment crawling: Web API querying and HTML data extraction. When the target Web page is integrated with a third-party commenting system whose Web API that is in Comquest’s knowledge base, it retrieves comments by sending HTTP requests to the API’s URL with parameters extracted from the target webpage. The approach has several challenges. Firstly, extracting accurate parameter values to construct HTTP requests is difficult since they are buried deep within the HTML string of web documents (if they exist). Secondly, the solution needs to generalize both vertically (within a website) and horizontally (across unseen websites). To tackle these challenges, the parameter extraction problem is treated as a variant of the multiclass Named Entity Recognition (NER) problem, where the entities represent the values of the parameters. Comquest leverages a sequential labeling deep learning model to identify parameter values within HTML source codes. When the commenting system is native to the website or unknown, Comquest detects and extracts user comments from fully rendered Web pages. However, comments are often hidden until triggered by specific user interaction, such as clicking on a designated page element among many other clickable elements. Furthermore, comments are typically presented as structured record-like Web data with high structure variations, making them difficult to detect and extract from the target Web page along with other record-like Web data. Comquest utilizes deep learning models and Web record extraction algorithms to automate the process of triggering, extracting, and classifying comments. Comquest has been implemented as a comprehensive system that consists of an administration web portal, a task controller, and a crawler backend. It provides a useful tool for collecting comments that represent a wider range of opinions, stances, and sentiments from websites on a global scale. / Computer and Information Science

The Qweak experiment has precisely determined the weak charge of the proton Qp w by measuring the parity-violating asymmetry in elastic electron-proton scattering at a low momentum transfer of Q2 = 0.0249 (GeV/c)2 . Qpw has a definite prediction in the Standard Model, and a value of sin2 θW can be extracted from it for comparison with other neutral current observables. Qweak measured the weak charge of the proton to be Qpw(P V ES) = 0.0719 ± 0.0045, which is consistent with the Standard Model value of Qp w(SM) = 0.0708 ± 0.0003. Qweak ran at the Thomas Jefferson National Accelerator Facility for two and a half years and was installed in experimental Hall C. A 180µA beam of longitudinally polarized electrons at 1.16 GeV scattered off a liquid hydrogen target of unpolarized protons. The electrons were collimated to an acceptance of 5.8◦ to 11.6◦ and then passed through a magnetic spectrometer and onto quartz Čerenkov detector bars. A detailed description of the theory and motivation behind the Qweak experiment is given. An overview of the Qweak apparatus and an in-depth discussion of the luminosity monitor performance is presented. A general overview of the Qweak analysis is also presented, with a focus on the beamline background correction, the nonlinearity measurement, and the simulation to constrain error for a rescattering effect. Also detailed here is the final, unblinded Qweak result, which determined Qpw to 6.2% and provided the highest precision measurement of sin2θW at low energy. / PHD / Q_weak is a precision-frontier accelerator driven experiment that took place at Thomas Jefferson National Accelerator Facility. Precision-frontier exists alongside the better known energy-frontier (which includes well known labs like the Large Hadron Collider) and refers to experiments which precisely measure values which are predicted by the latest theory. Deviations in these measurements help rule out theories and are used by energy-frontier experiments to know where to look for new physics. The Q_weak experiment measured the weak charge of the proton, which can be though of as the weak analog to electric charge. This value has never been measured before, and, since it is predicted to be small by current theory, is a good place to look for new physics. The value measured by this experiment indicates good agreement with the current theory. Even though there is good agreement with theory, Q_weak is an important result which will help define future physics models. In this thesis is an overview of the theoretical motivation of Q_weak, a general overview of the experimental design, in-depth discussion of the background detectors, general overview of the analysis with detailed descriptions of the several important corrections.