Computer Sciences Commons^™

Large Language Model Is Not A Good Few-Shot Information Extractor, But A Good Reranker For Hard Samples!, Yubo Ma, Yixin Cao, Yongchin Hong, Aixin Sun

Research Collection School Of Computing and Information Systems

Large Language Models (LLMs) have made remarkable strides in various tasks. However, whether they are competitive few-shot solvers for information extraction (IE) tasks and surpass fine-tuned small Pre-trained Language Models (SLMs) remains an open problem. This paper aims to provide a thorough answer to this problem, and moreover, to explore an approach towards effective and economical IE systems that combine the strengths of LLMs and SLMs. Through extensive experiments on nine datasets across four IE tasks, we show that LLMs are not effective few-shot information extractors in general, given their unsatisfactory performance in most settings and the high latency and …

Benchmarking Foundation Models With Language-Model-As-An-Examiner, Yushi Bai, Jiahao Ying, Yixin Cao, Xin Lv, Yuze He, Xiaozhi Wang, Jifan Yu, Kaisheng Zeng, Yijia Xiao, Haozhe Lyu, Jiayin Zhang, Juanzi Li, Lei Hou

Research Collection School Of Computing and Information Systems

Numerous benchmarks have been established to assess the performance of foundation models on open-ended question answering, which serves as a comprehensive test of a model’s ability to understand and generate language in a manner similar to humans. Most of these works focus on proposing new datasets, however, we see two main issues within previous benchmarking pipelines, namely testing leakage and evaluation automation. In this paper, we propose a novel benchmarking framework, Language-Model-as-an-Examiner, where the LM serves as a knowledgeable examiner that formulates questions based on its knowledge and evaluates responses in a reference-free manner. Our framework allows for effortless extensibility …

Molca: Molecular Graph-Language Modeling With Cross-Modal Projector And Uni-Modal Adapter, Zhiyuan Liu, Sihang Li, Yanchen Luo, Hao Fei, Yixin Cao, Kenji Kawaguchi, Xiang Wang, Tat-Seng Chua

Research Collection School Of Computing and Information Systems

Language Models (LMs) have demonstrated impressive molecule understanding ability on various 1D text-related tasks. However, they inherently lack 2D graph perception — a critical ability of human professionals in comprehending molecules’ topological structures. To bridge this gap, we propose MolCA: Molecular Graph-Language Modeling with Cross-Modal Projector and Uni-Modal Adapter. MolCA enables an LM (i.e., Galactica) to understand both text- and graph-based molecular contents via the cross-modal projector. Specifically, the cross-modal projector is implemented as a QFormer to connect a graph encoder’s representation space and an LM’s text space. Further, MolCA employs a uni-modal adapter (i.e., LoRA) for the LM’s efficient …

Neural Multi-Objective Combinatorial Optimization With Diversity Enhancement, Jinbiao Chen, Zizhen Zhang, Zhiguang Cao, Yaoxin Wu, Yining Ma, Te Ye, Jiahai Wang

Research Collection School Of Computing and Information Systems

Most of existing neural methods for multi-objective combinatorial optimization (MOCO) problems solely rely on decomposition, which often leads to repetitive solutions for the respective subproblems, thus a limited Pareto set. Beyond decomposition, we propose a novel neural heuristic with diversity enhancement (NHDE) to produce more Pareto solutions from two perspectives. On the one hand, to hinder duplicated solutions for different subproblems, we propose an indicator-enhanced deep reinforcement learning method to guide the model, and design a heterogeneous graph attention mechanism to capture the relations between the instance graph and the Pareto front graph. On the other hand, to excavate more …

Forecasting Traffic Speed During Daytime From Google Street View Images Using Deep Learning, Junfeng Jiao, Huihai Wang

Research Collection College of Integrative Studies

Traffic forecasting plays an important role in urban planning. Deep learning methods outperform traditional traffic flow forecasting models because of their ability to capture spatiotemporal characteristics of traffic conditions. However, these methods require high-quality historical traffic data, which can be both difficult to acquire and non-comprehensive, making it hard to predict traffic flows at the city scale. To resolve this problem, we implemented a deep learning method, SceneGCN, to forecast traffic speed at the city scale. The model involves two steps: firstly, scene features are extracted from Google Street View (GSV) images for each road segment using pretrained Resnet18 models. …

Anxiety In International Graduate Students With U.S. Education, Aeshah Zarraa

Theses and Dissertations

Anxiety in international graduate students is a significant concern for most students planning to move out of their home country to pursue higher education in the United States. The primary purpose of this research is to identify the prime causes of anxiety triggers in the targetted Graduate students, thereby determining a step-wise methodology development to address the causes. The study initially identified a set of graduate students who were voluntarily willing to collaborate and address their concerns anonymously to identify the significant issues faced by current students and alumni of the Florida Institute of Technology, Melbourne, Florida. Several questions were …

Responsibility Gaps And Black Box Healthcare Ai: Shared Responsibilization As A Solution, Benjamin H Lang, Sven Nyholm, Jennifer Blumenthal-Barby

Center for Medical Ethics and Health Policy Staff Publications

As sophisticated artificial intelligence software becomes more ubiquitously and more intimately integrated within domains of traditionally human endeavor, many are raising questions over how responsibility (be it moral, legal, or causal) can be understood for an AI’s actions or influence on an outcome. So called “responsibility gaps” occur whenever there exists an apparent chasm in the ordinary attribution of moral blame or responsibility when an AI automates physical or cognitive labor otherwise performed by human beings and commits an error. Healthcare administration is an industry ripe for responsibility gaps produced by these kinds of AI. The moral stakes of healthcare …

Foundations Of Memory Capacity In Models Of Neural Cognition, Chandradeep Chowdhury

Master's Theses

A central problem in neuroscience is to understand how memories are formed as a result of the activities of neurons. Valiant’s neuroidal model attempted to address this question by modeling the brain as a random graph and memories as subgraphs within that graph. However the question of memory capacity within that model has not been explored: how many memories can the brain hold? Valiant introduced the concept of interference between memories as the defining factor for capacity; excessive interference signals the model has reached capacity. Since then, exploration of capacity has been limited, but recent investigations have delved into the …

Random Variable Spaces: Mathematical Properties And An Extension To Programming Computable Functions, Mohammed Kurd-Misto

Computational and Data Sciences (PhD) Dissertations

This dissertation aims to extend the boundaries of Programming Computable Functions (PCF) by introducing a novel collection of categories referred to as Random Variable Spaces. Originating as a generalization of Quasi-Borel Spaces, Random Variable Spaces are rigorously defined as categories where objects are sets paired with a collection of random variables from an underlying measurable space. These spaces offer a theoretical foundation for extending PCF to natively handle stochastic elements.

The dissertation is structured into seven chapters that provide a multi-disciplinary background, from PCF and Measure Theory to Category Theory with special attention to Monads and the Giry Monad. The …

Hypothyroid Disease Analysis By Using Machine Learning, Sanjana Seelam

Electronic Theses, Projects, and Dissertations

Thyroid illness frequently manifests as hypothyroidism. It is evident that people with hypothyroidism are primarily female. Because the majority of people are unaware of the illness, it is quickly becoming more serious. It is crucial to catch it early on so that medical professionals can treat it more effectively and prevent it from getting worse. Machine learning illness prediction is a challenging task. Disease prediction is aided greatly by machine learning. Once more, unique feature selection strategies have made the process of disease assumption and prediction easier. To properly monitor and cure this illness, accurate detection is essential. In order …

Review Classification Using Natural Language Processing And Deep Learning, Brian Nazareth

Electronic Theses, Projects, and Dissertations

Sentiment Analysis is an ongoing research in the field of Natural Language Processing (NLP). In this project, I will evaluate my testing against an Amazon Reviews Dataset, which contains more than 100 thousand reviews from customers. This project classifies the reviews using three methods – using a sentiment score by comparing the words of the reviews based on every positive and negative word that appears in the text with the Opinion Lexicon dataset, by considering the text’s variating sentiment polarity scores with a Python library called TextBlob, and with the help of neural network training. I have created a neural …

Influence Of Pavement Conditions On Commercial Motor Vehicle Crashes, Stephen Arhin, Babin Manandhar, Adam Gatiba

Mineta Transportation Institute

Commercial motor vehicle (CMV) safety is a major concern in the United States, including the District of Columbia (DC), where CMVs make up 15% of traffic. This research uses a comprehensive approach, combining statistical analysis and machine learning techniques, to investigate the impact of road pavement conditions on CMV accidents. The study integrates traffic crash data from the Traffic Accident Reporting and Analysis Systems Version 2.0 (TARAS2) database with pavement condition data provided by the District Department of Transportation (DDOT). Data spanning from 2016 to 2020 was collected and analyzed, focusing on CMV routes in DC. The analysis employs binary …

Gr-397 Conceptualizing A Toc-Enhanced Chatbot: Pattern Recognition And Interaction, Sumaiya Tasneem, Sharon Elugoti, Chinni Cherrishma Reddy Aduri, Purna Pavan Kumar Kolli, Krishna Vamsi Anche

C-Day Computing Showcase

A chatbot is a software which is capable of communicating with human by using natural language processing. In our project, we plan to develop a Python-based chatbot that integrates theory of computation (TOC) concepts, including finite automata and regular expressions. The chatbot will interact with users, recognizing patterns and keywords in their inputs. We’ll begin by defining initial regular expressions for basic user interactions including greetings and inquiries.Future developments may enhance regular expressions and broaden the chatbot’s TOC-related capabilities, creating a versatile educational tool with practical TOC applications.

Gr-405 Boosting Clickbait Detection Through Semantic Insights And Attention-Driven Neural Network, Lokesh Meesala

C-Day Computing Showcase

The digital age has witnessed an explosion of online content, making it increasingly challenging for users to differentiate between reliable information and clickbait, which is often misleading or sensationalized. Clickbait contributes to the spread of misinformation, phishing attacks, and illegal marketing practices, and manipulates users’ decisions. Even from a business standpoint a clickbait might not lead to a conversion, A user might land on the page by following a clickbait and get frustrated and close the page. Additionally, with the increase in the usage of large language models for content writing it is even more challenging for the general user …

Gr-434 Phase Estimation’S Application In Qram, Ethan K Hunt

C-Day Computing Showcase

The paper proposes a new novel way of creating QRAM through quantum phase estimation. This is done by mapping a monotonically increasing sequence of natural numbers to a binary series and, ultimately, to a characteristic constant η which is then encoded as a phase in a quantum state. This process leverages quantum phase estimation, a fundamental quantum algorithm for finding the eigenvalues of a unitary operator which can be used as a form of QRAM in either Quantum or Hybrid models of computing

Gr-453 Medical Records Summarization Using Prompt-Based Nlp, Rawan Masadeh, Nicholas S Servies

C-Day Computing Showcase

In this paper, we present an innovative Natural Language Processing (NLP) algorithm for summarizing medical records extracted from the MIMIC-IV dataset using state-of-the-art (SOTA) techniques in text summarization. The increasing volume of electronic health records (EHRs) demands efficient methods for extracting meaningful insights from these complex and extensive documents. Our algorithm leverages recent advancements in NLP, including transformer-based models, to automate summarizing medical records while preserving critical information. Our algorithm is trained and tested using the Medical Information Mart for Intensive Care (MIMIC)-IV database that provides critical care data for over 40,000 patients admitted to intensive care units at the …

Egr-490 Importance Of Food Recognition On Blood Glucose Monitoring, Afnan Crystal

C-Day Computing Showcase

Maintaining blood sugar under control requires eating a healthy and balanced diet, exercising, and adhering to medications. Dietary consumption must be under strict control for diabetic patients’ general health. Traditional techniques for monitoring dietary consumption include recollection and manual record-keeping, which can be tedious and prone to mistakes. However, automated technologies for maintaining records that make use of computer vision, such as food image recognition systems, can streamline chronic health management for diabetics. These solutions seek to efficiently track daily food intake and consequential calories to facilitate and encourage lifestyle improvements. With this goal in mind, we design a Machine …

Gr-496 Cardiac Arrest Prediction Model, Vineeth Amsham, Sai Reddy Balaiah, Tamilkumar Subbarayakgounder

C-Day Computing Showcase

The "Cardiac arrest prediction model" project melds machine learning with healthcare to tackle heart disease. It aims to surpass current diagnostic tools that fail to catch early signs of cardiac events, often leading to high mortality. By developing an ML model that identifies early predictors of cardiac arrest, the project seeks to enable early interventions. Using supervised learning for its pattern recognition strength, the goal is to predict heart attacks accurately and thus, revolutionize preventative care and outcomes. This effort marks a leap in medical diagnostics and moves towards personalized healthcare, potentially saving countless lives and pioneering a new direction …

Gc-444 It Course Profile Website, Manikanta Voruganti

C-Day Computing Showcase

Build a Dynamic course profile website for Bachelor of science in information technology courses, that display all the information regarding the course.

Gc-448 Project Title: Discover, Learn, And Protect: A Mobile App For Informal Stem Learning About Local Biodiversity And Environmental Issues., Elvin Mccray, David Y. Appah, Adedunmola Banu, Binh Tran, Zenya Tucker

C-Day Computing Showcase

Our team assignment for this project was to create a mobile application that offers informal STEM learning about local biodiversity and environmental issues. Dr. Ying Xie, Professor in the College of Computing and Software Engineering (CCSE) is the owner of this project, who also laid out required features and provided necessary information, guidance, and advice for the project development. The core function of this application is to empower users to explore, identify and gain insights into the plant and animal species native to their region. Leveraging the capabilities of their smartphone’s camera, users can effortlessly scan, record, or locate local …

Gc-511 Predicting Stock Prices Using Different Machine Learning And Deep Learning Models, Afnan Crystal, Rohith Sundar Jonnalagadda, Hrithik Singh Chandel

C-Day Computing Showcase

Our project focuses on the challenge of predicting the daily closing prices and stock movements of Amazon, one of the world's largest and most dynamic corporations. Amazon's stock prices are known for their unpredictability and are influenced by a multitude of intricate factors. Our project aims to provide accurate and reliable forecasts for Amazon's stock prices, going beyond mere predictions. The analysis employs a comprehensive approach, comparing the performance of three distinct machine learning and deep learning models: Linear Regression, Support Vector Machine (SVM), and Multi-Layered Perceptron (MLP) for financial time series data. The dataset we used spans from January …

Gr-469 A Simulation Model Of The Traffic Signal System Using Java, Lingtao Chen

C-Day Computing Showcase

A traffic signal controls the flow of traffic at the intersection of two or more roadways. The first system of traffic signals was installed in London, England, in 1868. In this project, I will develop a simulation model for the traffic signal system using Java. The model will simulate the traffic signal system at a single four-way intersection. Also, I will compare the system performance with different input parameters, such as the number of vehicles and the cycle length, using various performance metrics, such as average waiting time and average sojourn time.

Euc-472 Biomimetic Remote-Controlled Vehicle, Jonathan Ridley, Kian Brown

C-Day Computing Showcase

The goal of this project is smoothly integrating instinctual concepts of control into devices beyond the body. It is essentially an attempt to extend the body without any complex prior training. To do this, we have developed a both a Bluetooth connection between hand movements and the motors of a multifaceted vehicle. Furthermore, the hand movements will be tracked using both an accelerometer and gyroscope found in the common hobbyist tool Arduino nano. Logging this data and processing it through the Bluetooth communication system, the intention is to provide real-time updates to the vehicle’s motors that ultimately sync the intentions …

Uc-491 Spectrum Analysis Cli Tool, Vitor B. Santos, Trey Redden, Sam Corella, Christopher Flores Santos, Jonathan Glennon

C-Day Computing Showcase

The Spectrum Analysis CLI Tool takes in .mp4 recordings of a Spectrum Analyzer, converts them programmatically into values the application can understand and outputs this data into a .csv file. This file can be parsed/filtered by the user with commands during upload of the .mp4 recording, or anytime after the recording has been processed.

Ur-510 Exploring The Impact Of Wavelength In Non-Invasive Blood Glucose Monitoring, John E. Oakley, Tahsin Kazi

C-Day Computing Showcase

Diabetes and metabolic diseases are some of the most crucial health issues of the 21st century. Monitoring blood glucose, the lead indicator of these diseases is a cumbersome process of constantly drawing blood or using subcutaneous needles. However, new technologies have emerged for non-invasive blood glucose monitoring that uses spectroscopy, which involves emitting light and capturing patient data with cameras. These new devices remove the cost of multiple tests, reduce the risk of skin conditions, and create more patient-friendly solutions. However, the hardware variables of these devices have not been tested thoroughly. One such avenue is via laser wavelength, which …

Gc-417 It Curriculum Success Portal, Alexander Thacker, Cedric Boakye-Danquah, Damola Adeyemo, Abdoulaye Ndiaye, Lydia Asante

C-Day Computing Showcase

In this project, we built a curriculum and course web portal to have all curriculum and course information in one place with easy search and browse interfaces. Currently course data and information are scattered in various places. These include essential information like course description, learning outcomes, sample syllabus, offering schedule and history. It also includes curriculum development information for department use, such as coordinator, developer, revision schedule, open learning materials. We collected all sources of IT course information and built a database to integrate the data in one place. Through this, we can build a complete profile of a course …

Gr-515 Developing A Conversational Chatbot Using Seq2seq Model With Tensorflow, Drashtee Parmar, Ruthvik R. Anugu

C-Day Computing Showcase

Sequence-to-Sequence (Seq2Seq) modeling, when paired with Long-Short-Term Memory (LSTM) units, has demonstrated significant potential in developing conversational chatbot capable of participating in text-based conversation and providing human-like responses.The Cornell Movie-Dialogs Corpus will be used to extract dialogues, preprocess the data, and then use the output to train the Seq2Seq model. Our contributions include exploring the application of LSTM for Natural Language Generation (NLG) and creating a comprehensive chatbot system. According to the results of the experiment, our method works well for coming up with thoughtful answers during a conversation.

Ur-409 Enhancing Aircraft Electronic Warfare Testing With Automated Rf Spectrum Analysis, Anthony De Santiago, Matthew T. Morgan, Geonhyeong Kim, Jalon L. Bailey, Camille Reaves

C-Day Computing Showcase

Military test ranges utilize a variety of Radio Frequency (RF) threat systems, to assess the effectiveness of Electronic Warfare (EW) systems during flight tests. A component of this process involves monitoring RF transmissions. Traditionally, system engineers at Robins Airforce Base have manually analyzed video from spectrum analyzers to confirm properties of specific threat systems. To streamline this analysis, our team's aim was to develop an automated solution for RF spectrum analysis. We employed a custom YOLO V8 model to isolate the analyzer screen and used a novel combination of frame differencing, summing, and agglomerative clustering techniques to extract relevant properties …

Eur-443 The Compression Connection: Ncd And Knn In Law Enforcement Text Analytics, Gabriel T. Gillott

C-Day Computing Showcase

Facing a deluge of digital records, law enforcement needs advanced data sorting systems. This project uses a new NLP model, blending compression algorithms and KNN, to categorize Cobb County police reports by mental health, behavioral, and drug issues—vital for efficient resource allocation. The model employs Normalized Compression Distance (NCD) to discern text similarities, enhancing analysis of varying report styles. Early tests show promise in label categorization, but generalizing remains challenging, marking future research directions. This NLP advancement could revolutionize data handling in public safety, aiming to surpass current classification standards.

Gc-412 Ecoedconnect, Vidhi Dave, Mythili Jayaraman, Manikanta Reddy Anugu, Rohini Paithanker, Neharika Beeram

C-Day Computing Showcase

An inventive educational platform called EcoEdConnect provides high school students with various opportunities to investigate biodiversity and environmental issues. By adjusting to each user's needs and choices, the web app offers a customized educational experience such as quizzes, experiments, videos, blogs, articles, etc. The project's first analysis, methodology, and early conclusions are presented in this document. It shows the several phases of the project, such as the introduction modules, practical experiments, discussions, blog, final assessment, and presentation, among other things. The application customizes the material and complexity according to the user's inclinations. Students' knowledge of biodiversity and environmental issues and …