Digital Commons Network^™

Solving Cold Start Problems Of Recommender Systems Using Generative Models, Panjawat Rungtranont

Chulalongkorn University Theses and Dissertations (Chula ETD)

Recommender systems are essential for enhancing user experiences on digital platforms, but they often struggle with suggesting new items lacking interaction data, known as the cold-start item problem. This study introduces "LLM-search," a novel method combining advanced language models (PaLM-2) with data augmentation techniques to address this challenge. LLM-search analyzes users' interactions to identify new items related to their preferences. It then employs a search engine to find cold items based on these preferences, effectively creating synthetic cold items interaction for training the recommender model. An experiment using the 2018 Amazon review dataset across sports & outdoors, groceries, and electronics …

Champions For Social Good: How Can We Discover Social Sentiment And Attitude-Driven Patterns In Prosocial Communication?, Raghava Rao Mukkamala, Robert J. Kauffman, Helle Zinner Henriksen

Research Collection School Of Computing and Information Systems

The UN High Commissioner on Refugees (UNHCR) is pursuing a social media strategy to inform people about displaced populations and refugee emergencies. It is actively engaging public figures to increase awareness through its prosocial communications and improve social informedness and support for policy changes in its services. We studied the Twitter communications of UNHCR social media champions and investigated their role as high-profile influencers. In this study, we offer a design science research and data analytics framework and propositions based on the social informedness theory we propose in this paper to assess communication about UNHCR’s mission. Two variables—refugee-emergency and champion …

Natural Language Processing In The Legal Domain, Daniel Martin Katz, Dirk Hartung, Lauritz Gerlach, Abhik Jana, Michael J. Ii Bommarito

Research Collection Yong Pung How School Of Law

In this paper, we summarize the current state of the field of NLP and Law with a specific focus on recent technical and substantive developments. To support our analysis, we construct and analyze a corpus of more than six hundred NLP and Law related papers published over the past decade. Our analysis highlights several major trends. Namely, we document an increasing number of papers written, tasks undertaken, and languages covered over the course of the past decade. We observe an increase in the sophistication of the methods which researchers deployed in this applied context. Slowly but surely, Legal NLP is …

A Novel Approach To Detecting And Mitigating Keyloggers, Damilola Osedumbi Elelegwu

College of Graduate Studies: Theses & Dissertations

As the digital world gets increasingly ingrained in our daily lives, cyberattacks—especially those involving malware—are growing more complex and common, which calls for developing innovative safeguards. Keylogger spyware, which combines keylogging and spyware functionalities, is one of the most insidious types of cyberattacks. This malicious software stealthily monitors and records user keystrokes, amassing sensitive data, such as passwords and confidential personal information, which can then be exploited. This research work introduces a novel browser extension designed to thwart keylogger spyware attacks effectively. The extension is underpinned by a cutting-edge algorithm that meticulously analyzes input-related processes, promptly identifying and flagging any …

Spoken Language To Irish Sign Language Machine Translation: A Linguistically Informed Approach, Jesus Aguilar Lopez, Irene Murtagh, Sheila Castilho

Academic Posters Collection

Many deaf and hard-of-hearing individuals rely on sign language (SL) on a daily basis as a preferred language. Although nowadays there are significant advances on spoken language research, current approaches are often neither linguistically motivated nor tailored to the unique features of SLs. Further research and development are necessary to enhance Sign Language Machine Translation (SLMT) and bring it to a similar level as spoken language MT. This research will endeavour to improve the accuracy and efficiency of SLMT systems, making them more accessible to the Deaf community and empowering deaf and hard of hearing individuals to communicate more effectively …

Towards Accommodating Gerunds Within The Sign Language Lexicon, Zaid Mohammed, Irene Murtagh

Academic Posters Collection

This work is part of ongoing research work that focuses on the linguistic analysis and computational description of five different Sign Languages (SLs) namely Irish Sign Language (ISL), Flemish Sign Language (VGT), Dutch Sign Language (NGT), Spanish Sign Language (LSE), and British Sign Language (BSL) as part of the SignON project. This work will be leveraged to inform the development of SL lexicon entries for a Sign Language Machine Translation (SLMT) system. In particular, this research focuses on ISL. We investigate the existence of constructions similar to or equivalent in functionality to gerunds in spoken language, in particular, English. The …

Neuemot: Mitigating Neutral Label And Reclassifying False Neutrals In The 2022 Fifa World Cup Via Low-Level Emotion, Ademola Adesokan, Sanjay Madria

Computer Science Faculty Research & Creative Works

Sports have been extensively studied for their impact on people's emotional well-being, with research revealing that they have the ability to reduce anxiety and unhappiness while boosting positive emotions1. among all sports, soccer stands out as the most popular and controversial2, eliciting a wide range of emotional reactions from fans, players, officials, and spectators, particularly on social media. While sentiment classifications such as positive, negative, and neutral have been extensively studied, low-level emotions, which refer to more specific and granular emotional states beyond the three basic categories, have yet to be given much attention. This study scraped over 300,000 tweets …

Robustembed: Robust Sentence Embeddings Using Self-Supervised Contrastive Pre-Training, Javad Asl, Eduardo Blanco, Daniel Takabi

School of Cybersecurity Faculty Publications

Pre-trained language models (PLMs) have demonstrated their exceptional performance across a wide range of natural language processing tasks. The utilization of PLM-based sentence embeddings enables the generation of contextual representations that capture rich semantic information. However, despite their success with unseen samples, current PLM-based representations suffer from poor robustness in adversarial scenarios. In this paper, we propose RobustEmbed, a self-supervised sentence embedding framework that enhances both generalization and robustness in various text representation tasks and against diverse adversarial attacks. By generating high-risk adversarial perturbations to promote higher invariance in the embedding space and leveraging the perturbation within a novel contrastive …

Deeppatent2: A Large-Scale Benchmarking Corpus For Technical Drawing Understanding, Kehinde Ajayi, Xin Wei, Martin Gryder, Winston Shields, Jian Wu, Shawn M. Jones, Michal Kucer, Diane Oyen

Computer Science Faculty Publications

Recent advances in computer vision (CV) and natural language processing have been driven by exploiting big data on practical applications. However, these research fields are still limited by the sheer volume, versatility, and diversity of the available datasets. CV tasks, such as image captioning, which has primarily been carried out on natural images, still struggle to produce accurate and meaningful captions on sketched images often included in scientific and technical documents. The advancement of other tasks such as 3D reconstruction from 2D images requires larger datasets with multiple viewpoints. We introduce DeepPatent2, a large-scale dataset, providing more than 2.7 million …

A Structure-Aware Generative Adversarial Network For Bilingual Lexicon Induction, Bocheng Han, Qian Tao, Lusi Li, Zhihao Xiong

Computer Science Faculty Publications

Bilingual lexicon induction (BLI) is the task of inducing word translations with a learned mapping function that aligns monolingual word embedding spaces in two different languages. However, most previous methods treat word embeddings as isolated entities and fail to jointly consider both the intra-space and inter-space topological relations between words. This limitation makes it challenging to align words from embedding spaces with distinct topological structures, especially when the assumption of isomorphism may not hold. To this end, we propose a novel approach called the Structure-Aware Generative Adversarial Network (SA-GAN) model to explicitly capture multiple topological structure information to achieve accurate …

Claimdistiller: Scientific Claim Extraction With Supervised Contrastive Learning, Xin Wei, Md Reshad Ul Hoque, Jian Wu, Jiang Li

Computer Science Faculty Publications

The growth of scientific papers in the past decades calls for effective claim extraction tools to automatically and accurately locate key claims from unstructured text. Such claims will benefit content-wise aggregated exploration of scientific knowledge beyond the metadata level. One challenge of building such a model is how to effectively use limited labeled training data. In this paper, we compared transfer learning and contrastive learning frameworks in terms of performance, time and training data size. We found contrastive learning has better performance at a lower cost of data across all models. Our contrastive-learning-based model ClaimDistiller has the highest performance, boosting …

Mining Health Informatics Job Advertisements: Insights For Higher Education Programs And Job Seekers, Ahmed El Noshokaty, Mohammad A. Al-Ramahi, Omar El-Gayar, Abdullah Wahbeh, Tareq Nasralah

Computer Information Systems Faculty Publications (Archived)

This paper used web scraping and data mining to analyze 831 health informatics job advertisements on indeed.com. Results showed that 87% of jobs explicitly required a college degree in a related field, 41% of jobs preferred a graduate degree, while 29% preferred or required professional certification. The analysis showed that preferred skills were analytics problem solving, communication skills, oral communication, interpersonal skills, project management, statistics, and critical thinking. The analysis also showed that college degrees, certifications, and the above-mentioned skill set are in high demand for working in the field of health informatics, especially in states with large populations and …

A Longitudinal Study Of Factors That Affect User Interactions With Social Media And Email Spam, Wojciech M. Mazurek

Graduate Theses, Dissertations, and Problem Reports (ETD)

Given the rapid growth of social media and the increasing prevalence of spam, it is crucial to understand users’ interactions with unsolicited content to develop effective countermeasures against spam. This thesis focuses on exploring the factors that influence users’ decisions to interact with spam on social media and email. It builds upon prior work, which serves as a foundation for further research and conducting a longitudinal analysis. Our results are based on the analysis of 221 responses collected through an online survey. The survey not only gathered demographic information such as age, gender, and race but also collected data on …

Tweetace: A Fine-Grained Classification Of Disaster Tweets Using Transformer Model, Ademola Adesokan, Sanjay Madria, Long Nguyen

Computer Science Faculty Research & Creative Works

Disaster management teams play a crucial role in responding to catastrophic events with speed and efficiency. However, when faced with large data of disaster-related information, manual systems can struggle to classify the information accurately, especially when they are unavailable. This challenge highlights the need for integrating social media and implementing machine learning models to address the issue. However, the development of such models is dependent on the availability of adequately annotated data, which presents a significant obstacle in the field of crisis management. to address this challenge, our study focuses on the need for disaster event classification through social media. …

Combating Fake News: A Gravity Well Simulation To Model Echo Chamber Formation In Social Media, Jeremy E. Thompson

Dartmouth College Ph.D Dissertations

Fake news has become a serious concern as distributing misinformation has become easier and more impactful. A solution is critically required. One solution is to ban fake news, but that approach could create more problems than it solves, and would also be problematic from the beginning, as it must first be identified to be banned. We initially propose a method to automatically recognize suspected fake news, and to provide news consumers with more information as to its veracity. We suggest that fake news is comprised of two components: premises and misleading content. Fake news can be condensed down to a …

Robots And Reference Services, Abdullahi Olayinka Isiaka, Biliamin Abiola Aremu, Abdulfatai Soliu, Fahisat Romoke Isiaq

Library Philosophy and Practice (e-journal)

Abstract

This paper explores the integration of robots into reference services in various library and information settings. The use of robots in these contexts has gained momentum in recent years, offering innovative solutions to enhance user experiences, improve access to information, and expand the capabilities of reference librarians. This paper reviews the current state of robots in reference services, discusses their potential benefits and challenges, and examines case studies to illustrate their practical applications. Furthermore, it offers insights into the future prospects and ethical considerations associated with the integration of robots in this domain. This paper delves into the fascinating …

Disciplines Different Of Fine-Grained Acknowledged Entity：A Case Of Doctoral Dissertations In Chinese Humanities And Social Sciences, Jiaxin He, Shijin Zhang, Jiayu Zheng, Xinlong Chu, Chengzhi Zhang

Journal of Scientific Information Research

[Purpose/significance]Thesis acknowledgment is a public text in which the authors express their gratitude to individuals and institutions those support their research. It is helpful to understand the individuals and institutions that play a role in the cultivation of doctoral candidate in different disciplines by extracting the entities of acknowledgment and comparing the distribution of the objects of acknowledgment in the dissertations of different disciplines.[Method/process]In this paper, we gained more than 60 000 Chinese doctoral dissertations acknowledgment of 21 humanities and social sciences disciplines, extracted acknowledged entities from acknowledgment, and constructed a fine-grained classification system of acknowledgment entities. The classification system …

Explainable Machine Learning For Evapotranspiration Prediction, Bamory Koné, Rima Grati, Bassem Bouaziz, Khouloud Boukadi

All Works

No abstract provided.

More Human Than Human: Llm-Generated Narratives Outperform Human-Llm Interleaved Narratives, Zoie Zhao, Sophie Song, Bridget Duah, Jamie C. Macbeth, Scott Carter, Monica Van, Nayeli Bravo, Matthew Klenk, Katherine Sieck, Alexandre Filipowicz

Computer Science: Faculty Publications

Narrative story generation has gained emerging interest in the field of large language models. The present paper aims to compare stories generated by an LLM only (non-interleaved) with those generated by interleaving human-generated and LLM-generated text (interleaved). The study’s hypothesis is that interleaved stories would perform better than non-interleaved stories. To verify this hypothesis, we conducted two tests with roughly 500 participants each. Participants were asked to rate stories of each type, including an overall score or preference and four facets—logical soundness, plausibility, understandability, and novelty. Our findings indicate that interleaved stories were in fact less preferred than non-interleaved stories. …

Image Schema Decompositions Of The Conceptual Dependency Ingest Primitive: A Study Of Paraphrases, Jamie C. Macbeth, Alexis Kilayko, Zoie Zhao, Sophie Song, Winniw X. Zheng

Computer Science: Faculty Publications

One of the hallmarks of the Schank-Minsky Conceptual Dependency Trans-Frames meaning representation system is that it attempts to express complex meanings by building large and complex conceptual structures using a relatively small number of primitives. Recently comparisons of image schemas with Conceptual Dependency primitives revealed ways of possibly reducing the number of primitives while maintaining the expressiveness of the set—an important research goal because it increases the flexibility and richness of the primitive-decomposed structures in a way that better approximates human cognition. Inspired by this prior work, we employ a paraphrase generation system to explore the replacement of the Conceptual …

Exploring The Evolution Of Ai Integration In English As A Foreign Language Education: A Scopus-Based Bibliometric Analysis (1997-2023), Mohanad G. Yaseen, Sara S. Alnakeeb

Mesopotamian Journal of Computer Science

Significant scholarly interest has been focused on the use of AI in English as a Foreign Language (EFL) instruction because it presents novel opportunities for improving language learning contexts and pedagogical approaches. This study provides an in-depth investigation into how AI is revolutionizing English as a foreign language (EFL) instruction, with special focus on the ways in which this technology can help students acquire a language in a fun, engaging, and one-on-one setting. This research examines the ethical and sociological implications of the broad integration of AI technology into language learning environments, highlighting major trends and identifying critical gaps through …

The Role Of Preprocessing For Word Representation Learning In Affective Tasks, Nastaran Babanejad, Heidar Davoudi, Ameeta Agrawal, Manos Papagelis

Computer Science Faculty Publications and Presentations

Affective tasks, including sentiment analysis, emotion classification, and sarcasm detection have drawn a lot of attention in recent years due to a broad range of useful applications in various domains. The main goal of affect detection tasks is to recognize states such as mood, sentiment, and emotions from textual data (e.g., news articles or product reviews). Despite the importance of utilizing preprocessing steps in different stages (i.e., word representation learning and building a classification model) of affect detection tasks, this topic has not been studied well. To that end, we explore whether applying various preprocessing methods (stemming, lemmatization, stopword removal, …

Examining Early Elementary Computer Science Identity Repertoires Within A Curriculum: Implications For Epistemologically Pluralistic Identities, Eleanor Richard, Shakhnoza Kayumova

Journal of Computer Science Integration

As computer science (CS) enters an increasing number of elementary classrooms, researchers must investigate the representations of what kinds of people are presented as doing computer science within CS curricula. In this paper, we explore a widely used, freely accessible, web-based, early elementary CS curriculum to examine the kinds of identity repertoires (behaviors, actions, skills, and socioemotional norms) that are promoted as representative of being/becoming a CS person. More specifically, we draw on identity studies and employ critical discourse analysis to examine how the kinds of norms and repertoires of CS practice made available in the curricular materials might construct …

Creating Data From Unstructured Text With Context Rule Assisted Machine Learning (Craml), Stephen Meisenbacher, Peter Norlander

School of Business: Faculty Publications and Other Works

Popular approaches to building data from unstructured text come with limitations, such as scalability, interpretability, replicability, and real-world applicability. These can be overcome with Context Rule Assisted Machine Learning (CRAML), a method and no-code suite of software tools that builds structured, labeled datasets which are accurate and reproducible. CRAML enables domain experts to access uncommon constructs within a document corpus in a low-resource, transparent, and flexible manner. CRAML produces document-level datasets for quantitative research and makes qualitative classification schemes scalable over large volumes of text. We demonstrate that the method is useful for bibliographic analysis, transparent analysis of proprietary data, …

Two Studies On Binary Program Stylometry: A White-Box Analysis And A Deep Learning Analysis On Some Crypto Ransomware, Alexander Mitchell

Masters Theses, 2020-current

Computer programmers often leave their individual programming styles in source code. Recent studies show that contrary to a popular belief, many of such programming styles can survive, in controlled environments, code compilation into binary. From the binary programming styles can be effectively retrieved for enhanced binary authorship attribution; such binary authorship attribution is often called binary program stylometry. In this thesis, we first perform a white-box impact analysis of various factors in code compilation on programming styles. For the MS Windows platform, we study the impact of multiple compilers, including gcc, Clang, and MSVC, their optimization levels, symbol stripping, and …

Low-Resource Machine Learning Techniques For The Analysis Of Online Social Media Textual Data, Toktam Amanzadeh Oghaz

Electronic Theses and Dissertations, 2020-2023

Low-resource and label-efficient machine learning methods can be described as the family of statistical and machine learning techniques that can achieve high performance without needing a substantial amount of labeled data. These methods include both unsupervised learning techniques, such as LDA, and supervised methods, such as active learning, each providing different benefits. Thus, this dissertation is devoted to the design and analysis of unsupervised and supervised techniques to provide solutions for the following problems: Unsupervised narrative summary extraction for social media content, Social media text classification with Active Learning (AL), Investigating restrictions and benefits of using Curriculum Learning (CL) for …

Full-Text Search Using Elasticsearch, Akash Shrestha

Masters Projects

Search engines have changed the way we use the internet. They can search or filter out relevant and valuable content of interest to the users. But many of the applications we use today lack search or are just poor. So how can we leverage the same power of search engines in our applications? This project aims to look at “Full-Text Search” which allows us to do a text-based search in text-intensive data. The search will be performed by matching any, or all words of the query exactly or with some relevancy against the indexes created by the searching tool. The …

Algorithmic Solutions To Combat Online Fake News, Xinyi Zhou

Dissertations - ALL

The unprecedented growth of new information producing, distributing, and consuming every moment on the Web has fostered the rise of ``fake news.'' Because of its detrimental effect on democracy, global economies, and public health, effectively combating online fake news has become an essential and urgent task.

This dissertation starts with making typological, theoretical, and empirical efforts to promote the public's comprehension of fake news and lay the foundation for algorithmically combating fake news. As there has been no universal definition of fake news, this dissertation discusses the definition of fake news from three dimensions: veracity, intention, and news, comparing it …

Iitd At The Wanlp 2022 Shared Task: Multilingual Multi-Granularity Network For Propaganda Detection, Shubham Mittal, Preslav Nakov

Natural Language Processing Faculty Publications

We present our system for the two subtasks of the shared task on propaganda detection in Arabic, part of WANLP'2022. Subtask 1 is a multi-label classification problem to find the propaganda techniques used in a given tweet. Our system for this task uses XLM-R to predict probabilities for the target tweet to use each of the techniques. In addition to finding the techniques, Subtask 2 further asks to identify the textual span for each instance of each technique that is present in the tweet; the task can be modeled as a sequence tagging problem. We use a multi-granularity network with …

Transfer Of Personality Through Text Style, Michael O'Mahony, Robert Ross

Other resources

The style of generated text is how something is said rather than what is said. We hypothesize that changing the style of generated text can change the perceived personality of the text generation agent. Dialogue systems that aim to imitate a human agent can appear to have a consistent personality through a consistent, controllable style of conversation. Some recent work on the style of generated text [1] performs impressively in the small number of domains selected for their experiments using transformer and LSTM-based models. Lin et al. [1] used weak supervised learning as their data set lacks parallel data. The …