Digital Commons Network^™

Learning To Generate Explainable Stock Predictions Using Self‑Reflective Large Language Models, Kelvin J.L. Koa, Yunshan Ma, Ritchie Ng, Tat‑Seng Chua

Research Collection School Of Computing and Information Systems

Explaining stock predictions is generally a difficult task for traditional non-generative deep learning models, where explanations are limited to visualizing the attention weights on important texts. Today, Large Language Models (LLMs) present a solution to this problem, given their known capabilities to generate human-readable explanations for their decision-making process. However, the task of stock prediction remains challenging for LLMs, as it requires the ability to weigh the varying impacts of chaotic social texts on stock prices. The problem gets progressively harder with the introduction of the explanation component, which requires LLMs to explain verbally why certain factors are more important …

Exploring Graph Neural Networks In Reinforcement Learning: A Comparative Study On Architectures For Locomotion Tasks, Gaukhar Nurbek

Theses and Dissertations

Deep Reinforcement learning (DRL) has gained importance in optimizing control policies, while Graph Neural Networks (GNNs) offer a robust approach for modeling complex relationships within systems represented as graphs. This thesis investigates the integration of DRL and GNNs to optimize control policies for robotic tasks, with a focus on locomotion. It compares static and dynamic GNN architectures for control policy predictions, revealing their strengths and limitations in adapting to locomotion predictions. The study assesses the impact of model structure complexity on GNNs' predictive capabilities, showcasing how intricate model structure can maximize GNNs' potential in capturing spatial and relational dependencies when …

Artificial Intelligence's Ability To Detect Online Predators, Olatilewa Osifeso

Electronic Theses, Projects, and Dissertations

Online child predators pose a danger to children who use the Internet. Children fall victim to online predators at an alarming rate, based on the data from the National Center of Missing and Exploited Children. When making online profiles and joining websites, you only need a name, an email and a password without identity verification. Studies have shown that online predators use a variety of methods and tools to manipulate and exploit children, such as blackmail, coercion, flattery, and deception. These issues have created an opportunity for skilled online predators to have fewer obstacles when it comes to contacting and …

An Empirical Study On The Efficacy Of Llm-Powered Chatbots In Basic Information Retrieval Tasks, Naja Faysal

Electronic Theses, Projects, and Dissertations

The rise of conversational user interfaces (CUIs) powered by large language models (LLMs) is transforming human-computer interaction. This study evaluates the efficacy of LLM-powered chatbots, trained on website data, compared to browsing websites for finding information about organizations across diverse sectors. A within-subjects experiment with 165 participants was conducted, involving similar information retrieval (IR) tasks using both websites (GUIs) and chatbots (CUIs). The research questions are: (Q1) Which interface helps users find information faster: LLM chatbots or websites? (Q2) Which interface helps users find more accurate information: LLM chatbots or websites?. The findings are: (Q1) Participants found information significantly faster …

Dynamic Storytelling Algorithms Using Contextual Aspects Of A Large Language Model, Alireza Pasha Nouri

Open Access Theses & Dissertations

Storytelling is a set of algorithms used to create narratives by connecting documents in a sequencethat accurately reflects the evolution of events and entities within a particular topic or theme. Early storytelling algorithms face challenges in encoding the progression and interconnections of information between consecutive texts, given that the conventional approaches rely primarily on connecting document pairs based on content overlap. They often neglect critical linguistic features, such as word contexts, semantics, the roles words play across different documents, and attention to the historical contexts of the underlying documents. Many existing storytelling models frequently produce story chains that, while connected …

Plug-And-Play Policy Planner For Large Language Model Powered Dialogue Agents, Yang Deng, Wenxuan Zhang, Wai Lam, See-Kiong Ng, Tat-Seng Chua

Research Collection School Of Computing and Information Systems

Proactive dialogues serve as a practical yet challenging dialogue problem in the era of large language models (LLMs), where the dialogue policy planning is the key to improving the proactivity of LLMs. Most existing studies enable the dialogue policy planning of LLMs using various prompting schemes or iteratively enhance this capability in handling the given case with verbal AI feedback. However, these approaches are either bounded by the policy planning capability of the frozen LLMs or hard to be transferred to new cases. In this work, we introduce a new dialogue policy planning paradigm to strategize LLMs for proactive dialogue …

Intriguing Properties Of Data Attribution On Diffusion Models, Xiaosen Zheng, Tianyu Pang, Chao Du, Jing Jiang, Min Lin

Research Collection School Of Computing and Information Systems

Data attribution seeks to trace model outputs back to training data. With the recent development of diffusion models, data attribution has become a desired module to properly assign valuations for high-quality or copyrighted training samples, ensuring that data contributors are fairly compensated or credited. Several theoretically motivated methods have been proposed to implement data attribution, in an effort to improve the trade-off between computational scalability and effectiveness. In this work, we conduct extensive experiments and ablation studies on attributing diffusion models, specifically focusing on DDPMs trained on CIFAR-10 and CelebA, as well as a Stable Diffusion model LoRA-finetuned on ArtBench. …

Reinforcement Retrieval Leveraging Fine-Grained Feedback For Fact Checking News Claims With Black-Box Llm, Xuan Zhang, Wei Gao

Research Collection School Of Computing and Information Systems

Retrieval-augmented language models have exhibited promising performance across various areas of natural language processing (NLP), including fact-critical tasks. However, due to the black-box nature of advanced large language models (LLMs) and the non-retrieval-oriented supervision signal of specific tasks, the training of retrieval model faces significant challenges under the setting of black-box LLM. We propose an approach leveraging Fine-grained Feedback with Reinforcement Retrieval (FFRR) to enhance fact-checking on news claims by using black-box LLM. FFRR adopts a two-level strategy to gather fine-grained feedback from the LLM, which serves as a reward for optimizing the retrieval policy, by rating the retrieved documents …

On The Feasibility Of Simple Transformer For Dynamic Graph Modeling, Yuxia Wu, Yuan Fang, Lizi Liao

Research Collection School Of Computing and Information Systems

Dynamic graph modeling is crucial for understanding complex structures in web graphs, spanning applications in social networks, recommender systems, and more. Most existing methods primarily emphasize structural dependencies and their temporal changes. However, these approaches often overlook detailed temporal aspects or struggle with long-term dependencies. Furthermore, many solutions overly complicate the process by emphasizing intricate module designs to capture dynamic evolutions. In this work, we harness the strength of the Transformer’s self-attention mechanism, known for adeptly handling long-range dependencies in sequence modeling. Our approach offers a simple Transformer model, called SimpleDyG, tailored for dynamic graph modeling without complex modifications. We …

Text-Attributed Graph Representation Learning : Methods, Applications, And Challenges, Ce Zhang, Menglin Yang, Rex Ying, Hady Wirawan Lauw

Research Collection School Of Computing and Information Systems

Text documents are usually connected in a graph structure, resulting in an important class of data named text-attributed graph, e.g., paper citation graph and Web page hyperlink graph. On the one hand, Graph Neural Networks (GNNs) consider text in each document as general vertex attribute and do not specifically deal with text data. On the other hand, Pre-trained Language Models (PLMs) and Topic Models (TMs) learn effective document embeddings. However, most models focus on text content in each single document only, ignoring link adjacency across documents. The above two challenges motivate the development of text-attributed graph representation learning, combining GNNs …

Automatic Grading Of Short Answers Using Large Language Models In Software Engineering Courses, Nguyen Binh Duong Ta, Yi Meng Chai

Research Collection School Of Computing and Information Systems

Short-answer based questions have been used widely due to their effectiveness in assessing whether the desired learning outcomes have been attained by students. However, due to their open-ended nature, many different answers could be considered entirely or partially correct for the same question. In the context of computer science and software engineering courses where the enrolment has been increasing recently, manual grading of short-answer questions is a time-consuming and tedious process for instructors. In software engineering courses, assessments concern not just coding but many other aspects of software development such as system analysis, architecture design, software processes and operation methodologies …

A Study Of Cybersecurity Landscape In The United States: Trend, Regional Variations, And Socioeconomic Factors, Trang Thi Thu Horn

All-Inclusive List of Electronic Theses and Dissertations

The Internet has become an essential part of our daily lives. Cyberspace has emerged significantly with an enormous number of users, creating a lucrative hunting field for cybercriminals. The exponential growth of internet users and online activities, along with the increased sophistication of cybercrimes, is raising significant global concerns for individuals, organizations, and governments. This research aims to explore the cybersecurity landscape in the United States, investigate regional variations, and the potential impact of the COVID-19 pandemic on the number of cybercrimes. The study utilizes a mixed research method approach, including historical analysis and a Delphi study. Historical analysis studied …

Space Transformation For Open Set Recognition, Atefeh Mahdavi

Theses and Dissertations

Open Set Recognition (OSR) is about dealing with unknown situations that were not learned by the models during training. In OSR, only a limited number of known classes are available at the time of training the model and the possibility of unknown classes never seen at training time emerges in the test environment. In such a setting, the unknown classes and their risk should be considered in the algorithm. Such systems require not only to identify and discriminate instances that belong to the source domain (i.e., the seen known classes contained in the training dataset) but also to reject unknown …

Representation Learning For Generative Models With Applications To Healthcare, Astronautics, And Aviation, Van Minh Nguyen

Theses and Dissertations

This dissertation explores applications of representation learning and generative models to challenges in healthcare, astronautics, and aviation.

The first part investigates the use of Generative Adversarial Networks (GANs) to synthesize realistic electronic health record (EHR) data. An initial attempt at training a GAN on the MIMIC-IV dataset encountered stability and convergence issues, motivating a deeper study of 1-Lipschitz regularization techniques for Auxiliary Classifier GANs (AC-GANs). An extensive ablation study on the CIFAR-10 dataset found that Spectral Normalization is key for AC-GAN stability and performance, while Weight Clipping fails to converge without Spectral Normalization. Analysis of the training dynamics provided further …

Advancing Mobile Sensing In Dynamic Environments, Weichen Wang

Dartmouth College Ph.D Dissertations

This thesis presents a comprehensive exploration of enhancing mobile sensing capabilities to address various aspects of human behavior, mental health, personality, social functioning and beyond. We redesign the StudentLife app to improve its sensing efficiency and dependability, enabling support for multi-year-long studies. By adopting new app design, this study addresses the technical challenges of continuous sensing and enhances system robustness. The work is organized into several key studies that collectively aim to expand the scope of mobile sensing in diverse and complex environments.

The first study broadens the scope of mobile sensing to assess personality traits, exploring the potential of …

A Discussion On Estimation Of The Best Constant For Spherical Restriction Inequalities, Hongyi Liu

Mathematics, Statistics, and Computer Science Honors Projects

The restriction conjecture asks for a meaningful restriction of the Fourier transform of a function to a sufficiently curved lower dimensional manifold. It then conjectures certain size estimates for this restriction in terms of the size of the original function. It has been proven in 2 dimensions, but it is open in dimensions 3 and larger, and is an area of much recent active effort. In our study, instead of aiming to prove the restriction conjecture, we target understanding its worst-case scenarios within known estimates. Specifically, we investigate the extension operator applied to antipodally concentrating profiles, examining the ratio of …

Russian Verbs Of Sound’S Web-Scraping Results From The A.A. Zalizniak Grammatical Dictionary And The Russian National Corpus. Multi-Dimensional Scaling Techniques And Visualization Strategies, John Simmons, Irina V. Ivliyeva

Graduate Student Research & Creative Works

This project aims to enhance web extraction techniques pertaining to a specific lexical-semantic group of Russian verbs of sounds, which undergo semantic modifications at the word-formation level (affixation). Additionally, it seeks to organize search results in a manner conducive to linguistic research using Multi-Dimensional Scaling (MDS) techniques and novel visualization strategies.

The primary objective in this phase of the research was to gather, consolidate, analyze, and present a comprehensive index of all forms of verbs of sound sourced from the A.A. Zalizniak Grammatical Dictionary of the Russian language, hyperlink each verbal form in this index with the Russian National Corpus …

Ur-94 Emohydra: Multimodal Emotion Classification Using Heterogenous Modality Fusion, William A Stigall

C-Day Computing Showcase

Affective computing is a field of growing importance, as human society becomes more integrated with machines. Human feelings are both complex and multi-modal, expressed through various methods and nuances in behavior. In this work we introduce EmoHydra, a multi-modal model created through the fusion of three top-level models fine-tuned on text, vision, and speech respectively. Despite heterogenous heads performing well on the unseen data, as well as generalizing well to other benchmarks, logit concatenation proves to be ineffective at predicting Multimodal data, therefore we implement Multi-Head Attention as our fusion mechanism.

Gpr-16 Attention Driven Framework For Detecting Mental Illness Causes From Social Media, Abm Adnan Azmee, Dinesh Chowdary Attota, Francis E Nweke

C-Day Computing Showcase

Mental health is a critical aspect of our overall well-being. Mental illness refers to conditions that impact an individual's psychological state, resulting in considerable distress, and limitations in functioning day-to-day tasks. Due to the progress of technology, social media has merged as the platform, for individuals to share their thoughts and emotions. The psychological state of individuals can be accessed with the help of data from these platforms. However, it is challenging for conventional machine learning models to analyze the diverse linguistic contexts of social media data. In this work, we propose a novel attention-driven deep framework to overcome these …

A System Of Communication Between Two Computers Using Novel Frequency Shift Keying Techniques, Jared Reyes

Honors Thesis

Frequency shift keying (FSK) is an old but powerful form of modulation that powered much of the early modems of the 1960’s, and the author felt inspired to make his own version of audio binary FSK modulation. He researched the general history and legacy of the Bell 103, a modem using FSK that defined telecommunication for the next few decades. Using research of the most common English characters of recent emails to determine which English characters should have the shortest bit length, a novel character encoding standard was created using variable bit rate. In addition, he has created a modulation …

A Holistic And Collaborative Behavioral Health Detection Framework Using Sensitive Police Narratives, Martin Keagan Wynne Brown

Dissertations

Identifying behavioral health is paramount for law enforcement officers to provide appropriate follow-up community care. In the current practice, law enforcement offices manually identify these behavioral health cases to allow the designation of the relevant follow-up resources. Police reports generated by officers' response to 911 calls remain an untapped resource for identifying such incidents. Therefore, we advocate for the incorporation of manual annotations from experts, natural language processing (NLP), active learning, advanced machine learning, and ensemble techniques to detect behavioral health cases within police reports. In this dissertation, we develop tools and frameworks to automatically detect behavioral health cases from …

Comprehensive Question And Answer Generation With Llama 2, Matous Hybl

MS in Computer Science Theses

Since the introduction of transformers, large language models have proven capable in many natural language processing fields. However, existing systems still face challenges in generating high-quality extractive questions. Base models and public chatbots fall short if the question source or quantity are critical. Our contribution is a question and answer generator for generating comprehensive, extractive questions and answers. This approach includes fine-tuning a LLaMA 2 base model for answer extraction (AE) and question generation (QG). We evaluate the resulting system using common automated metrics and a manual evaluation. We find that our system is comparable to the latest research and …

Immersive Japanese Language Learning Web Application Using Spaced Repetition, Active Recall, And An Artificial Intelligent Conversational Chat Agent Both In Voice And In Text, Marc Butler

MS in Computer Science Project Reports

In the last two decades various human language learning applications, spaced repetition software, online dictionaries, and artificial intelligent chat agents have been developed. However, there is no solution to cohesively combine these technologies into a comprehensive language learning application including skills such as speaking, typing, listening, and reading. Our contribution is to provide an immersive language learning web application to the end user which combines spaced repetition, a study technique used to review information at systematic intervals, and active recall, the process of purposely retrieving information from memory during a review session, with an artificial intelligent conversational chat agent both …

Discourse- And Lesion-Based Aphasia Quotient Estimation Using Machine Learning, Nicholas Riccardi, Satvik Nelakuditi, Dirk B. Den Ouden, Chris Rorden, Julius Fridriksson, Rutvik H. Desai

Communication Sciences and Disorders Faculty Articles and Research

Discourse is a fundamentally important aspect of communication, and discourse production provides a wealth of information about linguistic ability. Aphasia commonly affects, in multiple ways, the ability to produce discourse. Comprehensive aphasia assessments such as the Western Aphasia Battery-Revised (WAB-R) are time- and resource-intensive. We examined whether discourse measures can be used to estimate WAB-R Aphasia Quotient (AQ), and whether this can serve as an ecologically valid, less resource-intensive measure. We used features extracted from discourse tasks using three AphasiaBank prompts involving expositional (picture description), story narrative, and procedural discourse. These features were used to train a machine learning model …

Choosing Optimal Supply Radius Of Transformer Substations (Tss) In Iraq’S Cities Using Geometric Programming With Neutrosophic Coefficients, Ahmed K. Essa, Montifort Blessings Andrew Mitungwi, Tuweh Prince Gadama, Ahmed A. Salama

Neutrosophic Systems with Applications

Numerous uncertainties exist in various electricity power system problems due to the size, complexity, geographical distribution, and influence of unforeseen events in these systems, making it difficult for traditional mathematics tools based on crisp set theory to have an impact on and solve many power system problems. As a new branch of mathematical uncertainty techniques, the neutrosophic expert systems approach has therefore emerged with the development of electric power systems and has proven successful when correctly linked. The expert typically uses ambiguous or neutrosophic language to describe their empirical knowledge, such as "very likely," "quite likely," "if x is large, …

Advancing Text Summarization And Classification: Deep Insights From Transformer-Based Statistical Learning, Kun Bu

USF Tampa Graduate Theses and Dissertations

Artificial Intelligence (AI) is a part of human's daily life nowadays. Machine Learning (ML) as one aspect from AI has been rapidly developing during the past two decades, especially from the statistical learning approaches, which emphasized the use of probability and statistics to model data, such as Support Vector Machines (SVMs) for classification and regression tasks to the ensemble learning techniques, such as Random Forest, Gradient Boosting Machine (GBM), and stacking. Ensemble learning has evolved into a pivotal concept in contemporary machine learning, empowering practitioners to amalgamate multiple models to enhance generalization, accuracy, and robustness. As the field of machine …

Rethinking Plagiarism In The Era Of Generative Ai, James Hutson

Faculty Scholarship

The emergence of generative artificial intelligence (AI) technologies, such as large language models (LLMs) like ChatGPT, has precipitated a paradigm shift in the realms of academic writing, plagiarism, and intellectual property. This article explores the evolving landscape of English composition courses, traditionally designed to develop critical thinking through writing. As AI becomes increasingly integrated into the academic sphere, it necessitates a reevaluation of originality in writing, the purpose of learning research and writing, and the frameworks governing intellectual property (IP) and plagiarism. The paper commences with a statistical analysis contrasting the actual use of LLMs in academic dishonesty with educator …

Visualizing Routes With Ai-Discovered Street-View Patterns, Tsung Heng Wu, Md Amiruzzaman, Ye Zhao, Deepshikha Bhati, Jing Yang

Computer Science Faculty Publications

Street-level visual appearances play an important role in studying social systems, such as understanding the built environment, driving routes, and associated social and economic factors. It has not been integrated into a typical geographical visualization interface (e.g., map services) for planning driving routes. In this article, we study this new visualization task with several new contributions. First, we experiment with a set of AI techniques and propose a solution of using semantic latent vectors for quantifying visual appearance features. Second, we calculate image similarities among a large set of street-view images and then discover spatial imagery patterns. Third, we integrate …

Lyraquist: Language Learning Via Music App, Vivian D'Souza, Siri Avula, Mahi Patel, Tanvi Singh, Ashley Bickham

Senior Theses

Lyraquist is a new language learning mobile app that encourages the practice of a foreign language through music. Language learners can connect their Spotify Premium account to Lyraquist to listen to music in their target languages and utilize tools such as translation and vocabulary lists to facilitate language practice. Through integration with Spotify, users can import and create Spotify playlists and search the service’s entire catalog. By combining daily listening habits with several tasks associated with language learning in one place, Lyraquist hopes to be a useful language learning tool.

Sentiment Analysis Of Online Healthy Community Based On Semantic Enhancement, Pu Han, Ye Dongyu

Journal of Scientific Information Research

[Purpose/significance]In order to make full use of the value of text dependent syntactic information and prior emotion knowledge in emotion analysis， a semantic enhanced online healthy community emotion analysis model was proposed. [Method/process]Firstly， feature vectors for pre-processed online health community data are generated by Word2Vec and BERT; then local and global information of online review text are extracted using TextCNN and BiLSTM respectively based on dual-channel idea; then sentiment knowledge and dependency grammar information are merged in graph attention networks for semantic enhancement; finally， dual-channel features are fused and perform online health community sentiment classification in fully connected layer. [Result/conclusion]The …