Physical Sciences and Mathematics Commons^™

Towards A Computational Model Of Narrative On Social Media, Anne Bailey

Dartmouth College Undergraduate Theses

This thesis describes a variety of approaches to developing a computational model of narrative on social media. Our goal is to use such a narrative model to identify efforts to manipulate public opinion on social media platforms like Twitter. We present a model in which narratives in a collection of tweets are represented as a graph. Elements from each tweet that are relevant to potential narratives are made into nodes in the graph; for this thesis, we populate graph nodes with tweets’ authors, hashtags, named entities (people, locations, organizations, etc.,), and moral foundations (central moral values framing the discussion). Two …

Machine Learning And The Network Analysis Of Ethereum Trading Data, Santosh Sivakumar

Dartmouth College Undergraduate Theses

Since their conception, cryptocurrencies have captured the public interest, motivating a growing body of research aimed at exploring blockchain-based transactions. This said, little work has been done to draw conclusions from transaction patterns, particularly in the realm of predicting cryptocurrency price movements. Moreover, research in the cryptocurrency sphere largely focuses on Bitcoin, paying little attention to Ethereum, Bitcoin's second-in-line with respect to market capitalization. In this paper, we construct hourly networks for a year of Ethereum transactions, using computed graph metrics as features in a series of machine learning models. We find that regression-based approaches to predicting Ether prices/price deltas …

Designing Narrative-Based Interfaces For Collective Action: A Case Study Using Amazon, Climate Change, And Consumer Behavior, Catherine Parnell

Dartmouth College Undergraduate Theses

Climate change is the most pressing issue facing future generations. Amongst expanses of the population there is a lack of collective action on environmental issues, as there is a large gap between awareness and behavior change. This study suggests persuasive design that utilizes a narrative framing as a solution to reduce barriers to engaging in issues of collective action. Through extensive need-finding studies to understand target users, this thesis uses online-shopping via Amazon as a context for arguing that narrative can support actionable change in behavior. The technical artifact resulting from this research is a developed chrome extension and web …

Torsh: Obfuscating Consumer Internet-Of-Things Traffic With A Collaborative Smart-Home Router Network, Adam Vandenbussche

Dartmouth College Undergraduate Theses

When consumers install Internet-connected "smart devices" in their homes, metadata arising from the communications between these devices and their cloud-based service providers enables adversaries privy to this traffic to profile users, even when adequate encryption is used. Internet service providers (ISPs) are one potential adversary privy to users’ incom- ing and outgoing Internet traffic and either currently use this insight to assemble and sell consumer advertising profiles or may in the future do so. With existing defenses against such profiling falling short of meeting user preferences and abilities, there is a need for a novel solution that empowers consumers to …

Counting And Sampling Small Structures In Graph And Hypergraph Data Streams, Themistoklis Haris

Dartmouth College Undergraduate Theses

In this thesis, we explore the problem of approximating the number of elementary substructures called simplices in large k-uniform hypergraphs. The hypergraphs are assumed to be too large to be stored in memory, so we adopt a data stream model, where the hypergraph is defined by a sequence of hyperedges.

First we propose an algorithm that (ε, δ)-estimates the number of simplices using O(m^1+1/k / T) bits of space. In addition, we prove that no constant-pass streaming algorithm can (ε, δ)- approximate the number of simplices using less than O( m ^1+1/k / T ) bits of space. Thus …

Identifying Optimal Course Structures Using Topic Models, Tehut Tesfaye Biru

Dartmouth College Undergraduate Theses

This research project investigates whether there exists an optimal way to structure topics in educational course content that results in higher levels of engagement among students. It is implemented by fitting topic models to transcripts of educational videos contained in the Khan Academy platform. The fitted models were used to extract topic trajectories across time for each video and subsequently clustered based on whether they have similar “shapes”. The differences in mean engagement metrics per cluster suggest that some course shapes are more palatable to students regardless of subject matter. Additionally, the topic trajectories suggest a constant progression of topics …

A Configurable Social Network For Running Irb-Approved Experiments, Mihovil Mandic

Dartmouth College Undergraduate Theses

Our world has never been more connected, and the size of the social media landscape draws a great deal of attention from academia. However, social networks are also a growing challenge for the Institutional Review Boards concerned with the subjects’ privacy. These networks contain a monumental variety of personal information of almost 4 billion people, allow for precise social profiling, and serve as a primary news source for many users. They are perfect environments for influence operations that are becoming difficult to defend against. Motivated to study online social influence via IRB-approved experiments, we designed and implemented a flexible, scalable, …

Exploring The Relationship Between Intrinsic Motivation And Receptivity To Mhealth Interventions, Sarah Hong

Dartmouth College Undergraduate Theses

Recent research in mHealth has shown the promise of Just-in-Time Adaptive Interventions (JITAIs). JITAIs aim to deliver the right type and amount of support at the right time. Choosing the right delivery time involves determining a user's state of receptivity, that is, the degree to which a user is willing to accept, process, and use the intervention provided.

Although past work on generic phone notifications has found evidence that users are more likely to respond to notifications with content they view as useful, there is no existing research on whether users' intrinsic motivation for the underlying topic of mHealth …

Deterring Intellectual Property Thieves: Algorithmic Generation Of Adversary-Aware Fake Knowledge Graphs, Snow Kang

Dartmouth College Undergraduate Theses

Publicly available estimates suggest that in the U.S. alone, IP theft costs our economy between $225 billion and $600 billion each year. In our paper, we propose combating IP theft by generating fake versions of technical documents. If an enterprise system has n fake documents for each real document, any IP thief must sift through an array of documents in an attempt to separate the original from a sea of fakes. This costs the attacker time and money - and inflicts pain and frustration on the part of its technical staff.

Leveraging a graph-theoretic approach, we created the Clique-FakeKG algorithm …

Fine-Grained Detection Of Hate Speech Using Bertoxic, Yakoob Khan

Dartmouth College Undergraduate Theses

This thesis describes our approach towards the fine-grained detection of hate speech using deep learning. We leverage the transformer encoder architecture to propose BERToxic, a system that fine-tunes a pre-trained BERT model to locate toxic text spans in a given text and utilizes additional post-processing steps to refine the prediction boundaries. The post-processing steps involve (1) labeling character offsets between consecutive toxic tokens as toxic and (2) assigning a toxic label to words that have at least one token labeled as toxic. Through experiments, we show that these two post-processing steps improve the performance of our model by 4.16% on …

Lexical Complexity Prediction With Assembly Models, Aadil Islam

Dartmouth College Undergraduate Theses

Tuning the complexity of one's writing is essential to presenting ideas in a logical, intuitive manner to audiences. This paper describes a system submitted by team BigGreen to LCP 2021 for predicting the lexical complexity of English words in a given context. We assemble a feature engineering-based model and a deep neural network model with an underlying Transformer architecture based on BERT. While BERT itself performs competitively, our feature engineering-based model helps in extreme cases, eg. separating instances of easy and neutral difficulty. Our handcrafted features comprise a breadth of lexical, semantic, syntactic, and novel phonetic measures. Visualizations of BERT …

Interpreting Attention-Based Models For Natural Language Processing, Steven J. Signorelli Jr

Dartmouth College Undergraduate Theses

Large pre-trained language models (PLMs) such as BERT and XLNet have revolutionized the field of natural language processing (NLP). The interesting thing is that they are pre- trained through unsupervised tasks, so there is a natural curiosity as to what linguistic knowledge these models have learned from only unlabeled data. Fortunately, these models’ architectures are based on self-attention mechanisms, which are naturally interpretable. As such, there is a growing body of work that uses attention to gain insight as to what linguistic knowledge is possessed by these models. Most attention-focused studies use BERT as their subject, and consequently the field …

Improving Existing Methods For Calculating Embodied Carbon Emissions In Trade Through Feature Discovery: An Information Theoretic Approach, Sam Morton

Dartmouth College Undergraduate Theses

The continued societal and ecological risks posed by climate change have spurred renewed interest in quantitative tools that can improve policy aimed at climate mitigation. In 2008, international trade accounted for up to 26\% of global anthropogenic emissions, and therefore trade has garnered increased attention from policymakers seeking carbon mitigation. The concept of embodied carbon emissions in trade (EET) quantifies overall carbon emitted in the production and transport of goods for the purposes of trade. EET in theory could prove an indispensable tool to climate-concerned policymakers, but current implementations and data availability limit EET calculation to annual snapshots that extend …

Exploring The Long Tail, Joseph H. Hajjar

Dartmouth College Undergraduate Theses

The migration of datasets online has created a near-infinite inventory for big name retailers such as Amazon and Netflix, giving rise to recommendation systems to assist users in navigating the massive catalog. This has also allowed for the possibility of retailers storing much less popular, uncommon items which would not appear in a more traditional brick-and-mortar setting due to the cost of storage. Nevertheless, previous work has highlighted the profit potential which lies in the so-called "long tail'' of niche, unpopular items. Unfortunately, due to the limited amount of data in this subset of the inventory, recommendation systems often struggle …

Physically Based Rendering Techniques To Visualize Thin-Film Smoothed Particle Hydrodynamics Fluid Simulations, Aditya H. Prasad

Dartmouth College Undergraduate Theses

This thesis introduces a methodology and workflow I developed to visualize smoothed hydrodynamic particle based simulations for the research paper ’Thin-Film Smoothed Particle Hydrodynamics Fluid’ (2021), that I co-authored. I introduce a physically based rendering model which allows point cloud simulation data representing thin film fluids and bubbles to be rendered in a photorealistic manner. This includes simulating the optic phenomenon of thin-film interference and rendering the resulting iridescent patterns. The key to the model lies in the implementation of a physically based surface shader that accounts for the interference of infinitely many internally reflected rays in its bidirectional surface …

The Discrete-Event Modeling Of Administrative Claims (Demac) System: Dynamically Modeling The U.S. Healthcare Delivery System With Medicare Claims Data To Improve End-Of-Life Care, Rachael Chacko

Dartmouth College Undergraduate Theses

The shift of the U.S. healthcare delivery system from the treatment of acute conditions to chronic diseases requires a new method of healthcare system analysis to properly assess end- of-life (EOL) quality throughout the country. In this paper, we propose the Discrete-Event Modeling of Administrative Claims (DEMAC) system, which relies on a hetero-functional graph theory and discrete event-driven framework to dynamically model EOL care on multiple levels. The heat map visualizations produced by the DEMAC system enable the elucidation of not only patient-specific EOL care but also broader treatment patterns among providers and hospitals. As a whole, the DEMAC system …

Impulse Method For Shallow Water Simulation, Evan Muscatel

Dartmouth College Undergraduate Theses

The Shallow Water Equations is a simple method to simulate fluid in real-time. As a real-time model, the SWE is an excellent candidate for use in video games. However, the model is not often used in most fluid simulations because it does not preserve vorticity well, and therefore does not look very realistic. We present an improvement on the Shallow Water Equations by using a gauge method to preserve the vorticity of the fluid. We add a variable called impulse !, which is only weakly coupled with the velocity " of the simulation. We show that using this impulse method, …

Object Manipulation With Modular Planar Tensegrity Robots, Maxine Perroni-Scharf

Dartmouth College Undergraduate Theses

This thesis explores the creation of a novel two-dimensional tensegrity-based mod- ular system. When individual planar modules are linked together, they form a larger tensegrity robot that can be used to achieve non-prehensile manipulation. The first half of this dissertation focuses on the study of preexisting types of tensegrity mod- ules and proposes different possible structures and arrangements of modules. The second half describes the construction and actuation of a modular 2D robot com- posed of planar three-bar tensegrity structures. We conclude that tensegrity modules are suitably adapted to object manipulation and propose a future extension of the modular 2D …

An Inside Vs. Outside Classification System For Wi-Fi Iot Devices, Paul Gralla

Dartmouth College Undergraduate Theses

We are entering an era in which Smart Devices are increasingly integrated into our daily lives. Everyday objects are gaining computational power to interact with their environments and communicate with each other and the world via the Internet. While the integration of such devices offers many potential benefits to their users, it also gives rise to a unique set of challenges. One of those challenges is to detect whether a device belongs to one’s own ecosystem, or to a neighbor – or represents an unexpected adversary. An important part of determining whether a device is friend or adversary is to …

Probabilistic Error Upper Bounds For Distributed Statistical Estimation, Matthew Jin

Dartmouth College Undergraduate Theses

The size of modern datasets has spurred interest in distributed statistical estimation. We consider a scenario in which randomly drawn data is spread across a set of machines, and the task is to provide an estimate for the location parameter from which the data was drawn. We provide a one-shot protocol for computing this estimate which generalizes results from Braverman et al. [2], which provides a protocol under the assumption that the distribution is Gaussian, as well as from Duchi et al. [4], which assumes that the distribution is supported on the compact set [−1,1]. Like that of Braverman et …

Towards Ryser's Conjecture: Bounds On The Cardinality Of Partitioned Intersecting Hypergraphs, Anna E. Dodson

Dartmouth College Undergraduate Theses

This work is motivated by the open conjecture concerning the size of a minimum vertex cover in a partitioned hypergraph. In an r-uniform r-partite hypergraph, the size of the minimum vertex cover C is conjectured to be related to the size of its maximum matching M by the relation (|C|<= (r-1)|M|). In fact it is not known whether this conjecture holds when |M| = 1. We consider r-partite hypergraphs with maximal matching size |M| = 1, and pose a novel algorithmic approach to finding a vertex cover of size (r - 1) in this case. We define a reactive hypergraph to be a back-and-forth algorithm for a hypergraph which chooses new edges in response to a choice of vertex cover, and prove that this algorithm terminates for all hypergraphs of orders r = 3 and 4. We introduce the idea of optimizing the size of the reactive hypergraph and find that the reactive hypergraph terminates for r = 5...20. We then consider the case where the intersection of any two edges is exactly 1. We prove bounds on the size of this 1-intersecting hypergraph and relate the 1-intersecting hypergraph maximization problem to mutually orthogonal Latin squares. We propose a generative algorithm for 1-intersecting hypergraphs of maximal size for prime powers r-1 = pd under the constraint pd+1 is also a prime power of the same form, and therefore pose a new generating algorithm for MOLS based upon intersecting hypergraphs. We prove this algorithm generates a valid set of mutually orthogonal Latin squares and prove the construction guarantees certain symmetric properties. We conclude that a conjecture by Lovasz, that the inequality in Ryser's Conjecture cannot be improved when (r-1) is a prime power, is correct for the 1-intersecting hypergraph of prime power orders.

Autonomous Eye Tracking In Octopus Bimaculoides, Mark Andrew Taylor

Dartmouth College Undergraduate Theses

The importance of the position of cephalopods, and particularly octopuses, as the most intelligent group of invertebrates is becoming increasingly appreciated by the neuroscience research community. Cephalopods are the most distantly related species to humans that possesses advanced cognitive abilities; as their intelligence evolved independently from vertebrates, comparative analyses reveal trends in the evolution of nervous systems and the foundations of intelligence itself. Vision is an especially important area of cephalopod cognition to research because cephalopods are predominantly visual creatures, like humans, and the rapid transduction of visual signals allows the inner-workings of octopus cognition to be revealed in real …

A Computational Approach To Analyzing And Detecting Trans-Exclusionary Radical Feminists (Terfs) On Twitter, Christina T. Lu

Dartmouth College Undergraduate Theses

Within the realm of abusive content detection for social media, little research has been conducted on the transphobic hate group known as trans-exclusionary radical feminists (TERFs). The community engages in harmful behaviors such as targeted harassment of transgender people on Twitter, and perpetuates transphobic rhetoric such as denial of trans existence under the guise of feminism. This thesis analyzes the network of the TERF community on Twitter, by discovering several sub-communities as well as modeling the topics of their tweets. We also introduce TERFSPOT, a classifier for predicting whether a Twitter user is a TERF or not, based on a …

Restoring Humanity To Those Dying Below: An Inquiry Concerning The Ethics Of Autonomous Weapons Systems, Juliette A. Pouchol

Dartmouth College Undergraduate Theses

Today, autonomous weapons systems promise to make war more precise and effective while removing the human component from the battlefield. With the improvement of deep learning and computer vision, machines will soon be able to navigate and search through contested environments, discriminate between targets, and engage appropriately. The memoirs of drone pilots point to the evolving psychological impact of killing caused by the increase in the amount of empathy and emotional connectedness that drone pilots develop towards their target during the intimate surveillance period. A war fought without "skin-in-the-game" enables drone pilots to become better moral agents and decreases the …

Push-Relabel Algorithms For Computing Perfect Matchings Of Regular Bipartite Multigraphs, Benjamin J. Coleman

Dartmouth College Undergraduate Theses

We seek to compute perfect matchings of a d-regular bipartite multigraph G = (V, E). If d is a power of 2, we can perform Euler decomposition, which recursively separates a graph of even degree into subgraphs of smaller degree. When d is not a power of 2, however, Euler decomposition eventually returns a subgraph of odd degree. At this point, we can manually remove a perfect matching to return the graph to even degree and continue Euler decomposition. In this paper, we explore push-relabel algorithms as potential solutions to removing a perfect matching from a regular bipartite multigraph. Empirical …

Predicting Influencer Virality On Twitter, Danah K. Han

Dartmouth College Undergraduate Theses

The ability to successfully predict virality on Twitter holds great potential as a resource for Twitter influencers, enabling the development of more sophisticated strategies for audience engagement, audience monetization, and information sharing. To our knowledge, focusing exclusively on tweets posted by influencers is a novel context for studying Twitter virality. We find, among feature categories traditionally considered in the literature, that combining categories covering a range of information performs better than models only incorporating individual feature categories. Moreover, our general predictive model, encompassing a range of feature categories, achieves a prediction accuracy of 68% for influencer virality. We also investigate …

Mining Academic Publications To Predict Automation, Elena A. Doty

Dartmouth College Undergraduate Theses

This paper proposes a novel framework of predicting future technological change. Using abstracts of academic publications available in the Microsoft Academic graph, co-occurrence matrices are generated to indicate how often occupation and technological terms are referenced together. This matrices are used in linear regression models to predict future co-occurrence of occupations and technologies with a relatively high degree of accuracy as measured through the mean squared error of the models. While this work is unable to link the co-occurrences found in academic publications to automation in the labor force due to a dearth of automation data, future work conducted when …

Learning Humor Through Ai: A Study On New Yorker's Cartoon Caption Contests Using Deep Learning, Ray Tianyu Li

Dartmouth College Undergraduate Theses

My research focuses on predicting a cartoon caption's wittiness using multi-modal deep learning models. Nowadays, deep learning is commonly used in image captioning tasks, during which the machine has to understand both natural languages and visual pictures. However, instead of aiming to describe a real-world scene accurately, my research seeks to train computers to learn humor inside both natural languages and visual images. Cartoons are the artistic medium that supposes to deliver visual humor, and their captions are also supposed to be interesting to add to the fun. Thus, I decided to use research on cartoons' captions to see if …

Information Network Navigation, Ryan W. Blankemeier

Dartmouth College Undergraduate Theses

In this paper, we develop an interactive system to navigate information networks as a space with geometry, assigning each node in the network to geographical coordinates, and with that the ability to navigate as if on a map. A map-based rendering of the network gives the user the ability to understand meta-relationships (i.e., non-link-based relationships) that exist in the dataset that are lost with a traditional web search and (hyper-)link navigation. This requires first being able to represent the information corpus in such a way as to enable a quantifiable notion of similarity between the information nodes. A t-SNE (t-distributed …

Vr-Notes: A Perspective-Based, Multimedia Annotation System In Virtual Reality, Justin Luo

Dartmouth College Undergraduate Theses

Virtual reality (VR) has begun to emerge as a new technology in the commercial and research space, and many people have begun to utilize VR technologies in their workflows. To improve user productivity in these scenarios, annotation systems in VR allow users to capture insights and observations while in VR sessions. In the digital, 3D world of VR, we can design annotation systems to take advantage of these capabilities to provide a richer annotation viewing experience. I propose VR-Notes, a design for a new annotation system in VR that focuses on capturing the annotator's perspective for both "doodle" annotations and …