AUB ScholarWorks (American Univ. of Beirut)
Not a member yet
33634 research outputs found
Sort by
Towards an Unbiased Classification of Chest X-Ray Images Using a RL Powered ACGAN Framework
Get PDFComputer-aided diagnosis systems are invaluable tools for healthcare providers given the overwhelming volume of medical data at their disposal. However, a significant challenge of these systems is the existence of bias in their diagnostic outcomes, particularly affecting certain protected groups who are more susceptible to receiving incorrect diagnoses.
In this thesis, we investigate bias mitigation strategies, leveraging the discriminator of auxiliary conditional generative adversarial networks (ACGANs) as well as reinforcement learning (RL) agents for the classification of chest X-ray images. Our research targets bias reduction by utilizing reward functions designed to diminish the true positivity rate disparity (TPR Gap) between males and females for the discriminator.
We explore the impact of a hierarchical label distribution-based reward, a novel approach that aims to further improve bias in the diagnostic process. Through extensive evaluation and comparison, we study the disparities in precision, recall, f1 score, and true positivity rates across various different approaches. Also, we highlight the efficiency of each strategy in achieving more equitable diagnostic outcomes. The results show improvement in the TPR Gap when using the proposed method compared to the original classifier
Fine-Grained Arabic Named Entity Recognition
No full textThis thesis presents an approach to improve Fine-Grained Arabic Named Entity Recognition (NER) using a both supervised and semi-supervised deep learning models, utilizing both labeled and semi-labeled data. This study is motivated by the need for computers to process and interpret natural language effectively. We review a few similar studies on semi-supervised NER using Arabic language models and propose to build a large and reliable training dataset for Fine-Grained Arabic NER,which is largely overlooked in the field of Arabic NLP. Hence, we experiment with several annotation platforms and choose Labelbox for its online accessibility, support for fine- grained labeling, and a cluster of useful tools. We follow the FIGER dataset standard for named entities and sub-entities, and plan to build our own dataset consisting of around 2,000 Arabic Wikipedia articles. Our proposed semi-supervised deep learning model aims to improve on existing models in fine-grained Arabic NER models
Assessing the Effects of Thymoquinone (TQ) alone and in Combination with Cisplatin (CDDP) on Ovarian Cancer Cells
No full textBackground: Ovarian cancer, the most lethal gynecologic malignancy, ranks among the top five causes of cancer-related deaths in women. The high mortality rate is attributed to late-stage diagnosis, the absence of effective public screening methods, and resistance to platinum-based treatments. Consequently, combination drug therapy emerges as a promising avenue in ovarian cancer management, aiming to augment therapeutic efficacy and circumvent resistance to conventional therapies. Thymoquinone (TQ), a phytochemical compound derived from the Nigella sativa herb, has demonstrated anti-tumorigenic and anti-proliferative properties across various cancer types on ovarian cancer specifically when combined with platinum agents such as cisplatin (CDDP). Therefore, this study seeks to explore the impact of TQ on ovarian cancer cells, both alone and in combination with CDDP. Methods: Two ovarian cancer cell lines OVCAR-420 and SKOV-3 were used in this study. A range of two-dimensional (2D) in vitro assays, including MTT, trypan blue assay, and wound healing, were used to assess cell proliferation, viability, and migration respectively. Additionally, a three-dimensional (3D) MatrigelTM-based cell culture assay was employed to evaluate stemness. Both TQ and CDDP were administered as monotherapies and in combination to ascertain their effectiveness. Results: Our results revealed that TQ, both alone and in combination with cisplatin, significantly and synergistically reduced the proliferation, viability, migration, and sphere growth of ovarian cancer cells. TQ specifically at 25μM for OVCAR-420 and 10μM for SKOV-3, significantly lowered the concentration of CDDP when used in combination enhancing the susceptibility of cancer cells to treatment. Conclusion: Our findings show a strong synergistic effect between TQ and CDDP on ovarian cancer cells. Further validation of these findings could pave the way for a more potent therapeutic strategy for women diagnosed with ovarian cancer
Roundtable 2023
No full textThis discussion is the fourth of AUB-NCC’s 2023 series of roundtables, titled “Navigating Climate Change and environmental activism in the MENA region: Challenges and Opportunities”. Roundtable discussion Organized on September 21st 2023 at the Asfari Institute Conference Hall in AUB.The series is co-designed with: Nature Conservation Center, Issam Fares Institute for Public Policy and International Affairs, Asfari Institute for Civil Society & Citizenship, Heinrich-Böll-Stiftung, Arab Reform Initiative.Panel: Erica Accari, Co-founder and manager of Turba Farm ; Amani Beainy, Legal Researcher and co-founder of the Save Bisri campaign ; Cynthia Chidiac, Senior researcher at Asfari Institute for Civil Society & Citizenship (moderator)This brief was published following the fourth roundtable of AUB-NCC's 2023 series, titled “Navigating Climate Change and Environmental Activism in the MENA Region: Challenges and Opportunities.” The panel discussion shed light on the daily injustices and intersectional challenges faced by rural women in the context of climate change. By intertwining the stories of two young voices from the field, the discussion provided insights on how to influence national policies and initiatives, ensuring that the complex challenges of gender and climate justice receive the attention they warrant in Lebanon’s evolving narrative
Your body is yours
No full textA children's book, created by AUB students from the Education Department, for the course EDUC218 as a final project..آدم هو صبي صغير يواجه مواقف متعددة حيث يتعين عليه اتخاذ قرارات بشأن من يمكنه لمس جسد
Reliability-Based Design of Rammed Earth Structures using Machine Learning Models
No full textRammed earth construction is becoming popular since it successfully integrates eco-friendliness, sustainability, and the possibility of cost-effective construction. Rammed earth is comprised of clay, gravel, sand, silt, and may or may not include a stabilizer like cement or lime. Since the mechanical properties of stabilized rammed earth vary with properties of the soil and the type and level of cementation, a great deal of experimental and theoretical research has been done on the subject. The rigorous design considerations found in structures constructed using conventional building materials (such as concrete or masonry) are absent from structures created with rammed earth.
The absence of rigorous design practices for rammed earth poses a critical need for a systematic evaluation of the uncertainties inherent in the material properties of rammed earth and their impact on the reliability levels that characterize the likelihood of failure of structures built with these mixtures. The objective of this thesis is to use a data-driven approach to utilize a large database of experimental data on the unconfined compressive strength (UCS) of stabilized rammed earth mixtures to develop machine learning (ML) models that can predict the UCS as a function of the characteristics of the rammed earth mixture. The ML models will then be used within a reliability-based design framework to quantify the reliability levels that are inherent in the design of structural members which are constructed with rammed earth, and then compare the reliability levels to those currently enforced in conventional design codes. The results of this study could be used to inform the reliability-based design of rammed earth construction and to offer valuable insights into the design of sustainable and efficient construction materials.
By combining traditional knowledge with modern methodologies, this thesis aims to empower architects, engineers, and builders to harness the potential of rammed earth construction while meeting contemporary standards of safety and durability
Municipalities in Local Climate Governance: The Case Study of Menjez (Lebanon)
Get PDFLocal governments have increasingly become recognized as key players in the climate discourse. Yet, in the context of weak decentralization, their jurisdiction and influence are undermined. In Lebanon, the existing gap between national and local levels of government hinders proper mainstreaming of climate action. Nevertheless, a number of municipalities across the country have taken steps to address their carbon emissions and adopt adaptation measures within their territories.
This thesis examines the case study of Menjez, a small village in the North of Akkar that has demonstrated municipal efforts in becoming a low-carbon resilient village. In 2014, Menjez was one of the first municipalities in Lebanon to join the Covenant of Mayors (CoM) – an EU network that supports cities and local governments in advancing their climate agenda – signaling its commitment to reducing its local carbon emissions. To understand how climate change issues are governed in this small village, I identified four main factors building on Bulkeley et al., (2009) and Hoppe et al., (2016) (Motivation, National climate governance policy, Municipal Governance and Membership in Transnational Municipal Networks (TMNs)), and accordingly developed an evaluation framework. Data for this thesis relies on semi-structured interviews with experts in the field of energy and climate policy, as well as members of the municipality. The framework can serve as a tool to examine local efforts in advancing climate action, in the context of a weakly decentralized system
How to Research Inaccessible and Under-Studied Areas: The Case of Hawija, Iraq
Get PDFWhat are the alternative methods the scholars have utilized amid field inaccessibility due to conflict and post-conflict contexts? What methodological and ethical challenges researchers should be cognizant of when adopting a distant approach? What methods can be adopted to investigate regulatory systems in Hawija from afar based on its local context? This study explored distant methods and proposed a methodology for my originally planned field research on Hawija that was voted down due to security concerns. I reviewed the methods adopted by the scholars encountering empirical infeasibility in the Middle East and North Africa. By categorizing their choices of methods into three groups, this thesis analyzed their associated methodological and ethical challenges. In addition, to develop a practical methodology for conducting the planned research from afar, I resorted to local NGO staff and researchers possessing research experience in Hawija to inspect local context and logistical challenges. By proposing a combined approach of “glocal” collaboration and online interview, this work made methodological contributions to study conflict and post-conflict contexts, and more specifically, to investigate legal pluralism in areas of limited statehood
المفهوم القرآني لتحرير المرأة: قراءة في النموذج النِسائي
No full textيتضمن مراجع بيبليوجرافية.المقال عبارة عن مساهمة متواضعة قد توقظ فينا حبّ التساؤل وتجديد الفكر في المفهوم القرآني لتحرير المرأة” من خلال قراءة النموذج النسائيّ المثالي البارز في السور القرآنيّة وآياتها"
Automatic Personality Detection Through Text: Predicting the Big Five Traits from Self-Narratives
Get PDFAutomatic personality detection from text has gained interest since researchers discovered that linguistic style can be an indicator of personality. However, accurate personality classification remains a challenging task, often lacking data and robust evaluation metrics. This thesis investigates the ability of various machine learning models to predict the Big Five personality traits from text. We evaluate our models using two datasets. The first is the existing Stream of Consciousness Essays (SoCE) dataset, containing essays written by college students about their thoughts. The second is our newly collected Behavioral Interview Data (BID), featuring an annotated corpus tailored for this research. This new dataset includes university students' responses to behavioral questions similar to those asked in job interviews. In our experiments with both datasets, we explore different Natural Language Processing (NLP) techniques, focusing particularly on the Generative Pre-trained Transformer (GPT), using various parameters and testing methods. We compare GPT’s performance with a wide range of traditional and deep learning classifiers, including the BERT base model. Our key findings indicate that our data provides stronger indicators for detecting the Big Five traits than the SoCE dataset. Among the models tested, GPT-based approaches, notably GPT-4 (the latest version of GPT), consistently outperformed other approaches in identifying all five traits, even without prior training on the datasets. Additionally, we observe that fine-tuning GPT enhances its performance, particularly with the SoCE dataset. While achieving accuracy and F1 scores that are comparable to those in related studies, our research offers a more reliable evaluation of model performance by employing the Area Under the Curve (AUC) score, a metric that is more robust against data imbalance and sensitive model parameters. Moreover, our work underscores the practical applications of these models in real-world contexts, like behavioral job interviews, providing valuable insights for future research and applications in this field