Digital Commons Network^™

แบบจำลองการจัดลำดับความสำคัญของความเสี่ยงของยูสเซอร์สตอรีสำหรับการจัดลำดับความสำคัญโดยใช้ตรรกะคลุมเครือ, เปรม สุนทรภาส

Chulalongkorn University Theses and Dissertations (Chula ETD)

ในกระบวนการพัฒนาซอฟต์แวร์แบบแอไจล์ ยูสเซอร์สตอรีเป็นองค์ประกอบสำคัญที่สะท้อนความต้องการของผู้ใช้ แต่หากยูสเซอร์สตอรีมีคุณภาพไม่เพียงพอ อาจก่อให้เกิดความเสี่ยงที่ส่งผลกระทบต่อความสำเร็จของโครงการอย่างมีนัยสำคัญ ดังนั้นงานวิจัยนี้จึงนำเสนอแบบจำลองการจัดลำดับความสำคัญของความเสี่ยงของยูสเซอร์สตอรีโดยใช้ตรรกะคลุมเครือ เพื่อช่วยให้สามารถจัดการกับความไม่แน่นอนในการตัดสินใจได้อย่างมีประสิทธิภาพ โดยเริ่มจากการนิยามความเสี่ยงของยูสเซอร์สตอรีจำนวน 12 ประเภท พร้อมทั้งกำหนดเกณฑ์การประเมินใน 2 ด้าน ได้แก่ ด้านความพยายามในการบริหารจัดการ และด้านข้อจำกัดในการบริหารจัดการ จากนั้นจึงประยุกต์ใช้ตรรกะคลุมเครือร่วมกับตัวแปรภาษาและฟังก์ชันความเป็นสมาชิกในการคำนวณค่าคะแนนความเสี่ยง และจัดลำดับความสำคัญผ่านแบบจำลองที่พัฒนาขึ้น ผลการประเมินการใช้งานแบบจำลองกับผู้มีประสบการณ์ในโครงการแอไจล์พบว่า แบบจำลองสามารถช่วยในการจำแนกความเสี่ยงได้อย่างเหมาะสมและมีความสอดคล้องกับความคิดเห็นของผู้ประเมิน ซึ่งผลการประเมินสะท้อนถึงความสมเหตุสมผลของแบบจำลองและความสามารถในการประยุกต์ใช้จริง ดังนั้นงานวิจัยนี้จึงเป็นประโยชน์ต่อทีมพัฒนาในการประเมินและจัดลำดับความเสี่ยงของยูสเซอร์สตอรีได้อย่างเป็นระบบ และเพิ่มประสิทธิภาพในการบริหารความเสี่ยงในโครงการพัฒนาซอฟต์แวร์แบบแอไจล์ได้ดียิ่งขึ้น

6. More Is Less?: Using Generative Ai For Idea Generation And Diversification In Early Writing Processes, Franziska Tsufim, Lainie Pomerleau

Teaching and Generative AI: Pedagogical Possibilities and Productive Tensions

As writing teachers, we are strong proponents of process writing. At the same time, we are aware that early process work, especially in a group setting, can be time consuming and anxiety-inducing. Students may also self-censor when sharing work with peers especially if they are not confident in their ideas. Drawing on the process of nominal, electronic brainstorming, we created two different prompts that allowed students to incorporate generative Al into their idea generation process. This first activity improves the efficiency of individual idea generation, while the second exercise helps increase student confidence in their ideas in collaborative brainstorming situations. …

Digital Resurrection Of Historical Figures: A Case Study On Mary Sibley Through Customized Chatgpt, James Hutson, Paul Huffman, Jeremiah Ratican

Faculty Scholarship

This study investigates the emerging realm of digital resurrection, focusing on Mary Sibley (1800–1878), the esteemed founder of Lindenwood University. The core objective was to demonstrate the capability of advanced artificial intelligence, specifically a customized version of ChatGPT, in revitalizing historical figures for educational and engagement purposes. By integrating comprehensive diaries from Sibley with Claude 2.0, the research utilized a substantial autobiographical dataset to develop a GPT beta version that replicates her distinct voice and tone. The incorporation of her official portrait and diaries into the GPT Builder was pivotal, creating an interactive platform that accurately reflects her perspectives on …

Analysis Of Speech Recognition Systems And Error Correction Approaches, Saki Imai

Honors Theses

Despite significant advances in automatic speech recognition (ASR) accuracy, challenges remain. Naturally occurring conversation often involves multiple overlapping speakers, of different ages, accents and genders, as well as noisy environments and suboptimal audio recording equipment, all of which reduce ASR accuracy. In this study, we evaluate the accuracy of state of the art open source ASR systems across diverse conversational speech datasets, examining the impact of audio and speaker characteristics on WER. We then explore the potential of ASR ensembling plus post-ASR correction methods to improve transcription accuracy. Our findings underscore the need for robust error correction techniques and of …

Improving Question Answering Retrieval System Through Multi-Result Ranking Model, Danupat Khamnuansin

Chulalongkorn University Theses and Dissertations (Chula ETD)

Recent trends in various industries involve integrating artificial intelligence (AI) systems to enhance operational efficiency. Among these advancements, AI-supported question-answering (QA) systems have gained significant attention. These systems typically employ a two-stage process, integrating QA system capabilities with Information Retrieval (IR) methods. The introduction of the Retrieval Question Answering (ReQA) has further refined this process, offering a more practical solution, aiming to improve real-world applicability. Considering the wide availability of diverse QA retrieval models, employing a combination of multiple systems presents as a viable solution. However, the approach to combining QA retrieval systems remains relatively limited. We propose a method …

Enhancing Multilingual Sentence Representation Learning For The Job Recruitment Domain, Napat Laosaengpha

Chulalongkorn University Theses and Dissertations (Chula ETD)

With the advancement in natural language processing (NLP), there has been significant development in multilingual pretraining sentence encoder. Typically, these pretraining models are trained on large-scale datasets that consist of general text data from various sources such as Wikipedia. However, the general proposed models aren't enough to understand contexts in such a domain-specific, especially in the job recruitment domain. It is due to its niche nature and the lack of readily available related information. To enhance the existing multilingual pretraining sentence encoder and mitigate the aforementioned problems, we first propose multi-task dual-encoder framework to improve the sentence encoder for general-purpose …

Using Ontological Methods To Compare Cybersecurity Maturity Model Certification 2.0 And Cobit 19, Aaron Marshall Ramey

CCAC Theses and Dissertations

Cybersecurity frameworks developed by a variety of organizations and implemented by a much larger collection of organizations differ in their focus and application. Whether designed by a private or government organization, the primary goal is to provide a framework to assess and reduce risk. The Department of Defense (DoD) has recently implemented the second version of the Cybersecurity Maturity Model Certification (CMMC 2.0). In some situations, compliance with CMMC 2.0 has already become mandatory for the Defense Industrial Base (DIB). Compliance will soon be required for all Large Businesses (LB) and Small Businesses (SB) within the DIB. While COBIT 19 …

Constructed Language (Conlang) Audio Honing, Ronald B. Oakes

CCAC Theses and Dissertations

An important aspect of a constructed language (conlang) is how it sounds when spoken. This dissertation designs and implements a tool to allow the user to hear how their conlang sounds when spoken. This tool will generate spoken language based on sample text in the constructed language. Further, it will enable the user to manipulate the phonetics of the language and hear how these changes impact the language in its spoken form.

This tool also allows users to assess the preferability of the language’s phonetics using Net Auditory Distance under the framework of Beats-and-Bindings Phonology. It was shown that the …

Natural Language Processing And Neurosymbolic Ai: The Role Of Neural Networks With Knowledge-Guided Symbolic Approaches, Emily Barnes, James Hutson

Faculty Scholarship

Neurosymbolic AI (NeSy AI) represents a groundbreaking approach in the realm of Natural Language Processing (NLP), merging the pattern recognition of neural networks with the structured reasoning of symbolic AI to address the complexities of human language. This study investigates the effectiveness of neurosymbolic AI in providing nuanced understanding and contextually relevant responses, driven by the need to overcome the limitations of existing models in handling complex linguistic tasks and abstract reasoning. Employing a hybrid methodology that combines multimodal contextual modeling with rule-governed inferences and memory activations, the research delves into specific applications like Named Entity Recognition (NER), where architectures …

Dementia Detection Using Transfer Learning From Recorded Speech, Gun Wirojburapa

Chulalongkorn University Theses and Dissertations (Chula ETD)

Dementia is an irreversible and progressive condition that affects cognitive impairment, and millions of people are affected. Early detection provides patients with premature treatment and improves their quality of life. Traditional methods for dementia detection, such as cognitive assessment and neurological analysis, are expensive and time-consuming. Therefore, this thesis aims to research alternative methods to detect dementia, such as speech analysis. This thesis performs dementia classification by using transfer learning and machine learning techniques. Samples in the dataset consist of English and Chinese speakers. This study found transfer learning and machine learning techniques able to classify the speakers between dementia …

Thai-English Supported Automatic Speech Recognition For Endoscopic Reporting, Arpanant Saeng-Xuto

Chulalongkorn University Theses and Dissertations (Chula ETD)

This thesis presents the automatic speech recognition (ASR) system for endoscopic reporting that supports Thai-English code-switching. During endoscopic procedures, gastroenterologists are required to use both hands to handle instruments, thereby complicating the real-time documentation of abnormal findings. While recent advances in speech recognition offer promising solutions, existing models face difficulties with Thai-English code-switching and tend to overfit when fine-tuned on limited datasets. To overcome these limitations, we propose an ASR model enhanced with the Mixture of Experts (MoE) technique to improve transcription accuracy. Furthermore, the Named Entity Recognition (NER) model extracts gastrointestinal (GI) terminology from the transcriptions and classifies its …

Enhancing Large Language Models For Thai Legal Chatbots, Supachoke Hanwiboonwat

Chulalongkorn University Theses and Dissertations (Chula ETD)

Currently, developing a Thai legal question-answering system for the general public is highly challenging due to the complex, difficult-to-understand language and the extensive content of legal codes. This research proposes a Thai legal question-answering system designed for the public, aiming to establish best practices for developing effective legal QA systems. To improve performance, we created our own Thai legal QA dataset and incorporated data from various sources. We conducted comparative experiments to identify the most suitable language model for Thai legal contexts, and fine-tuned the models with diverse datasets for enhanced capabilities in legal QA and legal examinations. Additionally, we …

Emotion Recognition Through Advanced Signal Fusion And Kolmogorov-Arnold Networks, Aung Myo Thant

Chulalongkorn University Theses and Dissertations (Chula ETD)

Emotion recognition is an important aspect of affective computing that has many applications including human computer interaction and mental health care. In this thesis, Kolmogorov-Arnold Networks (KANs) are explored for the purpose of multimodal emotion recognition from EEG (Electroencephalogram) and eye movement signals of the SEED-V dataset. Both modalities were represented by Differential Entropy (DE) features and a two-branch KAN architecture was used which composed of seven KAN variants, namely, Efficient KAN, Jacobi KAN, Hermite KAN, Fourier KAN, Chebyshev KAN, RBF KAN, and Wavelet KAN (with different wavelet types including DOG, Morlet, Mexican hat, and Meyer). The strengths of two …

Craftjai: Llm-Enhanced Generative Ai For Gamified Mood Journal, Kingrak Phairoh

Chulalongkorn University Theses and Dissertations (Chula ETD)

Mood journaling helps promote emotional self-awareness, but maintaining user motivation remains difficult. This study presents CRAFTJAI, a generative AI system that enhances journaling through emotionally aligned visual feedback. Instead of replacing diary entries, the system adds character-based imagery to support intrinsic motivation and symbolic emotional reflection. It consists of emotional feature extraction using a large language model, emotion-labeled dataset construction based on eight core moods, and LoRA fine-tuning of a diffusion-based text-to-image model to improve emotional fidelity and stylistic consistency. A web-based prototype was developed with Gradio for real-time evaluation. Drawing on gamification and self-determination theory, the system integrates emotional …

Integrating Generative Artificial Intelligence With Systems Architecting Diagram Creation: Advancement, Challenges, Opportunities And Future Perspectives, Cansu Yalim, Holly H. Handley

Engineering Management & Systems Engineering Faculty Publications

Generative AI (GenAI) serves as a powerful tool that can create a wide range of content, including but not limited to text, speech, images, code, videos, and 3D models. ChatGPT stands out as a particularly appealing Generative Pretrained Transformer (GPT) model that offers supplementary capabilities through GPTs and plugins. These extensions enable users to engage with the chatbot and improve its functionality, surpassing mere content generation. Our study delves into the potential of ChatGPT, specifically GPT-4, to expedite the creation of diagrams to support the system architecting process. To this end, we explored the use of ChatGPT's Diagrams Show Me …

Identifying Fact-Checks Helpful For Vetting Factual Claims, Theodora Toutountzi

Computer Science and Engineering Dissertations - Archive

Recent efforts to combat misinformation have increasingly focused on automating the fact-checking process. For fact-checking systems to be automated, they need to recognize claims worth checking, match them with previously fact-checked claims, and determine their truthfulness. We refer to the task of matching unvetted claims with previously fact-checked claims as claim-matching. To be precise, the task is defined as follows: given a factual claim, identifying the fact-checks from a repository that could be helpful, or partially helpful, for vetting the given claim. A solution to this task is useful in practice since claimants often repeat the same claims even if …

The Manifesto For Teaching And Learning In A Time Of Generative Ai: A Critical Collective Stance To Better Navigate The Future, Aras Bozkurt, Junhong Xiao, Robert Farrow, John Y. H. Bai, Chrissi Nerantzi, Stephanie Moore, Jon Dron, Christian M. Stracke, Lenandlar Singh, Helen Crompton, Apostolos Koutropoulos, Evgenii Terentev, Angelica Pazurek, Mark Nichols, Alexander M. Sidorkin, Eamon Costello, Steven Watson, Dónal Mulligan, Sarah Honeychurch, Charles B. Hodges, Mike Sharples, Andrew Swindell, Isak Frumin, Ahmed Tlili, Patricia J. Slagter Van Tryon, Melissa Bond, Maha Bali, Jing Leng, Kai Zhang, Mutlu Cukurnova, Thomas K. F. Chiu, Kyungmee Lee, Stefan Hrastinski, Manuel B. Garcia, Ramesh Chander Sharma, Bryan Alexander, Olaf Zawacki-Richter, Henk Huijser, Petar Jandrić, Chanjin Zheng, Peter Shea, Josep M. Duart, Chryssa Themeli, Anton Vorochkov, Sunagül Sani-Bozkurt, Robert L. Moore, Tutaleni Iita Asino

STEMPS Faculty Publications

This manifesto critically examines the unfolding integration of Generative AI (GenAI), chatbots, and algorithms into higher education, using a collective and thoughtful approach to navigate the future of teaching and learning. GenAI, while celebrated for its potential to personalize learning, enhance efficiency, and expand educational accessibility, is far from a neutral tool. Algorithms now shape human interaction, communication, and content creation, raising profound questions about human agency and biases and values embedded in their designs. As GenAI continues to evolve, we face critical challenges in maintaining human oversight, safeguarding equity, and facilitating meaningful, authentic learning experiences. This manifesto emphasizes that …

Efficiently Learning An Encoder That Classifies Token Replacements And Masked Permuted Network-Based Bigru Attention Classifier For Enhancing Sentiment Classification Of Scientific Text, Muhammad Inaam Ul Haq, Khalid Mahmood, Qianmu Li, Ashok Kumar Das, Sachin Shetty, Majid Hussain

VMASC Publications

The exponential growth of scientific literature in digital repositories poses challenges in interpreting complex attitudes within academic texts. Traditional sentiment analysis methods often struggle with nuanced word meanings due to contextual variations. To address this, we propose the Electra-MPNet-based BiGRU attention classifier that extracts the high-level semantic features from citation sentences using the combined strength of Electra and MPNet encoder layers. These features are then combined to extract long-range dependencies through a stacked BiGRU layer. A linear attention mechanism is imposed to estimate the attention weights and context vector which enables the model to selectively focus on relevant information. The …

Multiview Outlier Filtered Pediatric Heart Sound Classification, Sagnik Dakshit

Computer Science Faculty Publications and Presentations

The advancements in deep learning has generated a large-scale interest in development of black-box models for various use cases in different domains such as healthcare, in both at-home and critical setting for diagnosis and monitoring of various health conditions. The use of audio signals as a view for diagnosis is nascent and the success of deep learning models in ingesting multimedia data provides an opportunity for use as a diagnostic medium. For the widespread use of these decision support systems, it is prudent to develop high performing systems which require large quantities of data for training and low-cost method of …

การพัฒนาเมตริกสำหรับการประเมินการสรุปข้อความแบบแอบสแทร็กต์ทิฟโดยใช้การวิเคราะห์ทางศัพท์และทางความหมาย, ณภัทร สารวนางกูร

Chulalongkorn University Theses and Dissertations (Chula ETD)

การประเมินผลการสรุปข้อความยังคงเป็นเรื่องท้าทาย เนื่องจากตัววัดแบบดั้งเดิม เช่น ROUGE และ BLEU มุ่งเน้นไปที่ความเหมือนกันของคำ ซึ่งมักจะไม่สามารถจับความหมายเชิงลึกและความสอดคล้องของเนื้อหาได้อย่างครบถ้วน งานวิจัยนี้นำเสนอ CorefSemScore ซึ่งเป็นตัววัดเชิงประเมินแบบผสมผสานที่รวมการวิเคราะห์ การแก้ไขการอ้างอิงร่วม เข้ากับการประเมินความคล้ายคลึงกันเชิงความหมาย โดยผสานการใช้ ROUGE, BERTScore, และ Sentence-BERT เพื่อการประเมินที่ครอบคลุมมากขึ้น CorefSemScore ใช้วิธีการเฉลี่ยแบบถ่วงน้ำหนักเพื่อผสานองค์ประกอบการประเมินในหลากหลายมิติ โดยเน้นความสมดุลระหว่างการประเมินความคล้ายคลึงกันเชิงคำศัพท์และความคล้ายคลึงกันเชิงความหมาย ส่งผลให้สามารถประเมินการสรุปข้อความแบบแอบสแทร็กต์ทิฟได้อย่างครอบคลุม ผลการทดลองบนชุดข้อมูล SummEval แสดงให้เห็นว่า CorefSemScore มีประสิทธิผลเหนือกว่าทั้งตัววัดแบบความคล้ายคลึงกันเชิงคำศัพท์ เช่น ROUGE และ BLEU รวมถึงตัววัดความคล้ายคลึงกันเชิงความหมาย เช่น BERTScore โดยเฉพาะในด้านการสะท้อนความเชื่อมโยงของเนื้อหา ซึ่งสะท้อนถึงศักยภาพของตัววัดในการยกระดับความน่าเชื่อถือของการประเมินผลการสรุปข้อความแบบแอบสแทร็กต์ทิฟ

Language Models For Rare Disease Information Extraction: Empirical Insights And Model Comparisons, Shashank Gupta

Theses and Dissertations--Computer Science

End-to-end relation extraction (E2ERE) is a crucial task in natural language processing (NLP) that involves identifying and classifying semantic relationships between entities in text. This thesis compares three paradigms for end-to-end relation extraction (E2ERE) in biomedicine, focusing on rare diseases with discontinuous and nested entities. We evaluate Named Entity Recognition (NER) to Relation Extraction (RE) pipelines, sequence-to-sequence models, and generative pre-trained transformer (GPT) models using the RareDis information extraction dataset. Our findings indicate that pipeline models are the most effective, followed closely by sequence-to-sequence models. GPT models, despite having eight times as many parameters, perform worse than sequence-to-sequence models and …

Analysing Natural Language Processing Techniques: A Comparative Study Of Nltk, Spacy, Bert, And Distilbert On Customer Query Datasets., Patrizia De Camillis

ICT

This study examines the role of sentiment analysis in customer queries, emphasising its impact on brand perception and the risks of poor query management. It compares the performance of NLP models—NLTK, spaCy, BERT, and DistilBERT—on customer query and feedback data. The findings show that BERT and DistilBERT produce similar results, often categorising queries as neutral, indicating their strength in handling diverse sentiments. NLTK and spaCy also share performance patterns. The research offers insights into the capabilities and limitations of these models in sentiment analysis.

Advancing Discourse Analysis In Multiparty Meetings: Comprehensive Classification Of Argument And Relation Types, Vishal Vaitla

Master's Projects

In multi-party meetings, accurately analyzing dialogue is crucial for enhancing communication effectiveness and decision-making. However, the informal and dynamic nature of these discussions presents complex challenges for computational analysis. Dialogues in such settings often include non-standard language, interruptions, and rapid topic changes, making it difficult to extract useful information with conventional text analysis tools. To tackle this challenge, two specific methods were developed:

Argument Classification: We use machine learning models like Gradient Boosting to identify and categorize the main points people make in their discussions. This helps us understand what each person is trying to say, making it easier to …

Neurosymbolic Ai Approach To Attribution In Large Language Models, Deepa Tilwani, Revathy Venkataramanan, Amit P. Sheth

Faculty Publications

Attribution in large language models (LLMs) remains a significant challenge, particularly in ensuring the factual accuracy and reliability of the generated outputs. Current methods for citation or attribution, such as those employed by tools like Perplexity.ai and Bing Search-integrated LLMs, attempt to ground responses by providing real-time search results and citations. However, so far, these approaches suffer from issues such as hallucinations, biases, surface-level relevance matching, and the complexity of managing vast, unfiltered knowledge sources. While tools like Perplexity.ai dynamically integrate web-based information and citations, they often rely on inconsistent sources such as blog posts or unreliable sources, which limits …

Embedding Software Engineering In Mixed Methods: Computationally Enhanced Risk Communication, Ann Marie Reinhold, Madison H. Munro, Elizabeth A. Shanahan, Ross J. Gore, Barry C. Ezell, Clemente I. Izurieta

VMASC Publications

Mixed methods research ameliorates many convergent research challenges within the contemporary sociotechnical landscape. We suggest the integration of software engineering in mixed methods studies is a critical step to address some of the remaining and persistent challenges. One such research challenge where software engineering is particularly well suited is in hazard preparedness—in particular, the creation of risk communication messages to mitigate or prevent harm. Computationally enhanced risk communication is convergent research that integrates software engineering and social science research for the benefit of protecting humans and infrastructure. To this end, we developed a mixed methods framework for the efficient construction …

Building A Human Digital Twin (Hdtwin) Using Large Language Models For Cognitive Diagnosis: Algorithm Development And Validation, Gina Sprint, Maureen Schmitter-Edgecombe, Diane Cook

Computer Science Faculty Scholarship

Background: Human digital twins have the potential to change the practice of personalizing cognitive health diagnosis because these systems can integrate multiple sources of health information and influence into a unified model. Cognitive health is multifaceted, yet researchers and clinical professionals struggle to align diverse sources of information into a single model. Objective: This study aims to introduce a method called HDTwin, for unifying heterogeneous data using large language models. HDTwin is designed to predict cognitive diagnoses and offer explanations for its inferences. Methods: HDTwin integrates cognitive health data from multiple sources, including demographic, behavioral, ecological momentary assessment, n-back test, …

Social Networks And Large Language Models For Division I Basketball Game Winner Prediction, Gina Sprint

Computer Science Faculty Scholarship

Sporting event outcome prediction is a well-established and actively researched domain, with a particular focus on college basketball’s March Madness tournament. Researchers, fans, and gamblers alike seek accurate game-level predictions using features such as tournament seeds, season performance, and expert opinions. While machine learning algorithms have been harnessed to build prediction models, no perfect model or human-created bracket has emerged. This paper explores a novel approach to basketball game outcome prediction by utilizing the power of social networks and large language models (LLMs). LLMs are trained to understand and generate text, often eliminating the need for a feature engineering step. …

Deep Transfer Learning-Based Bird Species Classification Using Mel Spectrogram Images, Mrinal Kanti Baowaly, Bisnu Chandra Sarkar, Md.Abul Ala Walid, Md. Martuza Ahamad, Bikash Chandra Singh, Eduardo Silva Alvarado, Imran Ashraf, Md. Abdus Samad

School of Cybersecurity Faculty Publications

The classification of bird species is of significant importance in the field of ornithology, as it plays an important role in assessing and monitoring environmental dynamics, including habitat modifications, migratory behaviors, levels of pollution, and disease occurrences. Traditional methods of bird classification, such as visual identification, were time-intensive and required a high level of expertise. However, audio-based bird species classification is a promising approach that can be used to automate bird species identification. This study aims to establish an audio-based bird species classification system for 264 Eastern African bird species employing modified deep transfer learning. In particular, the pre-trained EfficientNet …

Integrating Knowledge Graphs With Large Language Models For Natural Language Querying, Rakesh Kandula

Browse all Theses and Dissertations

This research explores the integration of knowledge graphs with large language models that have already been trained on a vast pool of unstructured text data. Large language models trained on this type of data have a tendency to hallucinate and produce factually inaccurate results. This behavior is primarily due to the data being trained is unstructured and huge text corpus, and large language model uses predictive text analysis methods to obtain a response. These issues can be addressed by applying Retrieval Augmented Generation and Fine-tuning to large language models, employing an underlying domainspecific knowledge graph. Integrating knowledge graph and large …

A Comparison Of Lexical Tokenization Methods, Nathan Culmer

Williams Honors College, Honors Research Projects

The purpose of this project was to compare tokenization methods, or methods of breaking up a text into meaningful parts for use in natural language processing. The effectiveness of several commonly used tokenization methods were investigated, including morpheme tokenization, which takes into account the linguistic features of the language. In addition, I proposed and implemented a new technique to consider the capitalization pattern of a word in the tokenization process, in order to allow this process to include more natural language features. The effectiveness of these methods was compared by using them in a sentiment analysis model for various datasets, …